H ealth literacy and health outcomes in China’s floating population: mediating effects of health services

: Background: Health literacy is an important behavioral factor for promoting health and disease prevention. This study aimed to examine whether health literacy affected health outcomes in China’s floating population and whether health service utilization had a mediating effect between health literacy and health outcomes. Method: A cross-sectional study was carried out in Zhejiang Province, China, in November and December 2019. Self-reported questionnaires were used for data collection, which included sociodemographic characteristics, health literacy, health outcomes, and health service utilization. On the basis of reliability testing, confirmatory factor analysis was used to test questionnaire validity. Descriptive statistics were used to understand the demographic characteristics of the floating population, and structural equation modeling was used for the mediation test to check whether health service utilization had a mediating effect between health literacy and health outcomes. Results: There were positive correlations between health literacy, health service utilization, and health outcomes; correlation coefficients ranged from 0.165 to 0.944. Mediation analysis showed that health service utilization had partial mediating effects between health literacy and health outcomes. In the relationship between health literacy and health outcomes, the indirect effects of health service utilization accounted for 6.6%–8.7% of the total effects. Conclusion: Health service utilization has partial mediating effects between health literacy and health outcomes. Health literacy affects the proactiveness of health service utilization in the floating population through healthcare literacy and health promotion, thereby affecting health outcomes. and Family Planning Dynamic Monitoring Questionnaire is into four physical examination; five-point Likert scales to in these and promotion partial

it is important to focus on the health problems of this population.
Proposing that the individual is ultimately responsible for his or her health, the "Healthy China" strategy advocated increasing health literacy, shifting the focus from disease treatment to prevention, and accelerating the adoption of healthy lifestyles. The World Health Organization has defined health literacy as "the cognitive and social skills which determine the motivation and ability of individuals to gain access to, understand and use information in ways which promote and maintain good health" [8]. Studies have found that health literacy is associated with access to health services, and lower health literacy leads to adverse health behavior and outcomes [9, 10].
One study found that increased disease-prevention literacy in a Turkish rural population was positively correlated with vaccination rates [11]. Due to low health literacy, China's floating population is at a high risk for infectious diseases; in particular, a lack of reproductive health knowledge has increased the spread of sexually transmitted diseases [12]. Many youths in the floating population have irregular diets, excess stress, and a lack of routine medical treatment.
One study found a correlation between the health literacy level of Korean migrants and the risk factors for type 2 diabetes [13]. Moreover, poor health literacy and poor care were found to be correlated with HIV infection among the African American population [14]. Meanwhile, one study found that while locals and immigrants face similar mental health problems, immigrants are less likely to use mental health services [15]. Another study found that health literacy could pose a barrier to accessing health services among vulnerable populations, thereby resulting in poor health outcomes [16].
Health service utilization includes medical, healthcare, and rehabilitation service utilization.
One study found that maternal health services significantly reduced mortality rates among pregnant women [17]. Healthcare systems based on place of residence and employment status can make it difficult for migrants to access health services [18]. Due to the restrictions of China's household registration system, the floating population has low awareness of health services [19], poor utilization of health services, and low vaccination rates compared to local populations [20].
Outside of China, health service utilization by immigrants is similarly affected by health literacy.
In New York City, low mental health service utilization rates among elderly Chinese immigrants increased the risk of mental illness [21]. In Brazil, only 45.6% of immigrant subjects used some form of medical service [22]. In light of such findings, there is a need to examine how to improve health literacy among floating populations to promote health service utilization.
In summary, based on the abovementioned studies of both Chinese and international migrants, we should focus on three dimensions of health promotion to improve the health of the floating population-namely, health awareness, health behavior, and supportive environments. Health awareness and health behavior are part of health literacy while a supportive environment is related to providing various external health services. An individual's health literacy must merge with a supportive external environment to ultimately improve health outcomes. This study aimed to examine the relationship between health literacy and health outcomes in the floating population and determine whether health service utilization has a mediating effect in this relationship.

METHOD
The data used in this study were obtained from Zhejiang Province, China. Zhejiang is located at the southern part of the Yangtze River Delta and includes two subprovincial cities, nine prefecture-level cities, 35 counties, 21 county-level cities, and 34 urban districts. The "2018 China Floating Population Development Report" indicated that China's floating population was 8.01%, respectively, of the floating population. After the floating population in Zhejiang had exceeded that of Shanghai and Jiangsu in 1997 and 2000, respectively, Zhejiang had the second-largest floating population in China for 19 continuous years. This, along with the province's level of economic development, is why Zhejiang was selected for this study. For the study sites, we selected Hangzhou, Ningbo, and Wenzhou, which are representative cities with internal migrant aggregation in China. Before conducting the survey, a preliminary test was carried out in which 150 questionnaires were distributed and 148 were collected (recovery rate: 98%). Cronbach's alpha tests and confirmatory factor analysis were conducted on the questionnaire before revision. The revised questionnaire was then used to carry out a large-scale survey in November 2019.
The survey period was November to December 2019. The 2017 annual floating population data for Zhejiang Province were used as the sampling frame basis. A stratified, multistage, scale-proportion probability proportional to size (PPS) method was used for sampling. The top three cities with the largest floating populations in Zhejiang were selected for the survey (i.e., Hangzhou, Ningbo, and Wenzhou). Three sampling sites were randomly selected for each city, and three communities were randomly selected in the sampling site. Then, the researchers selected 20-40 people from the floating population in each selected community according to gender, age, and migration time. The inclusion criteria included people who had been in Zhejiang for at least one year before the survey, did not have a registered household address in the region (county, city), and were over age 15 in November 2019. The exclusion criteria included students and transient populations at train stations, harbors, airports, hotels, and hospitals. A total of 670 questionnaires were distributed online and offline, and 657 valid questionnaires were recovered (validity rate: 98%). Before the subject completed the self-reported questionnaire, the investigator explained the aim of study, the data collection method, and how to complete the survey. The investigators also informed subjects that participation was anonymous and voluntary.

Health gain
The dependent variables of self-evaluated health, physiological health, and psychological

Health literacy
The European Health Literacy Survey Questionnaire was used to investigate health literacy in the floating population. This questionnaire is multidimensional and has been used to measure health literacy in European populations [27]. The questionnaire is available in three versions.
First, the HLS-EU-Q47 is based on a 4 (information processing: finding, understanding, judging, applying) × 3 (health domain: health care, disease prevention, health promotion) matrix and contains 47 items. Second, the HLS-EU-Q86 adds health behavior, health status, health service utilization, community participation, sociodemographic, and socioeconomic factors and has 86 questions. Third, the HLS-EU-Q16 was formulated for rapid health evaluation. Among these three, the HLS-EU-Q47 is the most widely used, and it has high validity and reliability in Asian contexts [28]. Therefore, the HLS-EU-Q47 questionnaire was used in this study. It includes 47 questions measuring health literacy scored using four-point Likert scales (1: extremely difficult to 4: extremely easy). The lowest possible mean score is 1, and the highest is 4.

Data analysis
SPSS 22.0 was used for frequency analysis, reliability testing, and Pearson's correlation analysis. Amos 22.0 was used to establish a confirmatory factor model to provide validity, construct a standardized path test, and examine the hypothesis testing results; bootstrapping was used to test the mediating effects. In the reliability analysis, the baseline value for judging the questionnaire was determined to be 0.7, with a value greater than 0.7 indicating that the questionnaire was feasible. Similarly, the Amos test of questionnaire validity was carried out using the model fit index. The specific criteria for assessment were cmin/df<5 and GFI, AGFI, NFI, TLI, and CFI<0.8, showing that the questionnaire had good validity. If a model has three questions, the constructed model is saturated and df is 0, and the model fit results will not be evaluated. In subsequent hypothesis testing, the collinearity test results were first used to ensure the inflation factor of the variable did not exceed 10 to show that collinearity was absent between variables. At the same time, the common-method variance test was carried out. Harman's single-factor test was used to determine that the explanatory power of the first component in the initial eigenvalue was 42.972%, which is lower than 50%, showing that common-method variance was absent between variables. After that, path modeling and mediator model testing were conducted. The basis of the path modeling was the theoretical model ( Figure 1). Figure 1 shows that healthcare, disease prevention, and health promotion were the independent variables; health service utilization was a mediator; and health gain was the dependent variable; these were used to construct the path model. In this study, the path significance was p < 0.05, showing that the predicted relationship between the variables was proven. At the same time, bootstrapping was used for the mediator test to obtain the total, direct, and mediating effects of various variables. A total of 2,000 random sampling calculations were used to obtain the 95% confidence interval of the estimated value. If the confidence interval contains 0, the test results are not true; otherwise, the test results are true. Table 1 shows the sociodemographic characteristics of the study participants. Among the 657 subjects, there were 244 males (37.1%) and 413 females (62.9%). The proportion of subjects aged <40 years was 75.6%, and 78.2% had an educational level of high school or below. The proportion of subjects with a bachelor's degree was 9.4%. Table 2 shows the reliability analysis of the scale and the variable means. Cronbach's α consistency coefficient was used to test internal consistency reliability, and the threshold value was 0.7. A Cronbach's α greater than 0.7 means the question measurement results for that dimension had consistent reliability. The results show that the reliability of various sections was greater than 0.7, indicating that the questionnaire was reliable. The mean health literacy value showed that the health promotion literacy of the floating population was better than their healthcare literacy and disease-prevention literacy. The mean health service utilization value showed that health examination service utilization was better than other types of health services.

RESULTS
The mean value for health outcomes showed that the overall health status of the floating population was poor, and the psychological health results were better than the physiological health results, which validates the findings of previous studies. Table 3 shows the validity analysis results for the scale. Using Amos for structural equation modeling (SEM), the evaluation markers included CMIN/DF, NFI, IFI, TLI, CFI, RMSEA, and other fit markers. These markers were mainly used for model discrimination. Table 3 shows the specific discrimination criteria. From the confirmatory factor analyses of the three questionnaires, it can be seen that the model fit indices for health literacy, health service utilization, and health outcomes were within the acceptable range. Since health outcomes has only three questions and is a saturated model, a model fit test was not carried out. From this, it can be determined that the various questionnaires had good validity markers and that the questionnaires are valid. Table 4 shows the correlation relationships between variables. The total health service utilization score and the various dimensions showed a significantly positive correlation with the total score for health literacy and various dimensions. At the same time, the total health service utilization score and the various dimensions showed a significantly positive correlation with the total score for health outcomes and various variables. Moreover, the total score for health literacy and various dimensions showed a significantly positive correlation with the total health gain score and various variables. The overall model fit results were good, with CMIN/DF=1.723< 3; NFL, IFI, TLI, and CFI were all greater than 0.9, and RMSEA=0.033<0.05. Therefore, it can be determined that the data and the model match very well, and the model is valid.
Bootstrapping was used for the mediating effect test with 2,000 samples. The 95% confidence interval test was used to obtain the mediating effect results. The mediation test was significant for healthcare effects on health outcomes through health service utilization, with a total effect size of 0.280; the confidence interval [0.174-0.378] does not include 0, indicating that the total effect is true. The direct effect size was 0.215, and the confidence interval

DISCUSSION
This study employed SEM to examine the relationship between health literacy and health outcomes in China's floating population and simultaneously analyzed the mediating effects of health service utilization. Health literacy affected health outcomes through two dimensions-healthcare literacy and health promotion literacy-but disease prevention did not directly affect health outcomes. The results showed that health literacy had positive effects on health outcomes, and health service utilization had partial mediating effects between health literacy and health outcomes.
Healthcare literacy directly affected health outcomes. Healthcare literacy concerns whether people are equipped with an understanding of and ability to communicate about health services and whether they possess medical skills for emergencies. The proportion of the floating population who chose to self-medicate or not adopt any measures was the highest, followed by visiting private/individual clinics or community health service stations/rural health centers. The proportion of subjects who chose to go to private/individual clinics was 1/5, and the proportion of subjects who chose to go to county-level and above hospitals was the lowest [29]. Such choices are the result of interactions between subjective and objective factors and are closely associated with education level, income, occupation, health status, and disease. Self-medication was the first choice among the floating population after disease onset, which requires actively searching for treatment information. If they are unable to self-medicate, they usually suffer from a major disease and have to seek professional medical services. This requires them to be able to communicate with physicians and understand medication instructions, among other aspects of treatment. At the same time, income and social security limit the ability of the floating population to access medical services. They must therefore master certain self-treatment techniques, such as traditional massage, acupuncture, and cupping therapy.
Health promotion literacy directly affected health outcomes. Health promotion refers to a social behavior or strategy that uses administrative or organizational measures to coordinate various social departments, communities, families, and individuals to carry out individuals' health responsibilities to jointly maintain and promote health. In health literacy, health promotion refers to an individual's understanding of the factors affecting physical and mental health, such as governmental health policies, community facilities, social networks, work environments, and residential environments. It also refers to actively searching for relevant health education Disease prevention did not directly affect health outcomes. Disease-prevention health literacy mainly refers to understanding the importance of health behaviors and health examinations. It entails understanding the effects of smoking, low exercise intensity, and excessive drinking; knowing that vaccinations and health examinations can help prevent disease; and possessing the ability to improve health behaviors. Workers should also be able to decide how to protect themselves from disease based on suggestions from friends, family, and media. In this regard, this study's results conflict with previous findings. It could be that the floating population is one that is naturally selected for health, and, generally, only healthy people can migrate. Meanwhile, most of China's floating population were originally farmers, and their low educational levels limit their understanding of nutrition, healthcare, and disease prevention. Most floating population workers do manual labor. Under profit maximization and industry competition pressure, small and medium-sized enterprises may choose to increase manufacturing speed to produce more products. They employ a piece-rate payment system to encourage workers to work longer hours. This not only affects their health but also reduces normal rest and leisure time, which includes time for exercise and attention to diet. Long working hours, poor living conditions, poor work environments, and stress related to integration cause them to lack the energy to gain disease prevention knowledge and cultivate healthy lifestyle habits.
Health service utilization directly affected health outcomes. In 2016, China proposed the "Healthy China" strategy, establishing public health as a primary objective. However, data from Lastly, health impact evaluation is a typical task in public health services. A study in Germany found that health assessment can clarify the responsibilities of health services [38]. Therefore, assessing the impact of various types of health services on health for resource matching is a future option.
Health service utilization had partial mediating effects in the relationship between healthcare literacy and health promotion literacy and health outcomes. Health service utilization was not found to have a mediating effect in the relationship between disease prevention literacy and health outcomes. The effects of health literacy on health outcomes can be partially explained by health service utilization. For example, females with lower health literacy tend to use less preventive healthcare services, including flu vaccination and cervical and breast cancer screening [39]. Moreover, low health literacy is more common among elderly people [40]. A lack of health literacy also directly affects the effective utilization of health services and social welfare by chronic disease patients and is closely associated with disease management and health outcomes [41]. Therefore, good health literacy in the floating population can enable them to clearly understand, recognize, and control the relationship between their lifestyle and health status [42].
Improving health knowledge, changing unhealthy lifestyle and work habits (e.g., working overtime or working when sick), actively using various available health resources, and making a habit of regular physical exams can form a positive feedback loop of early diagnosis, treatment, and recovery. In summary, the floating population poses a top health priority, and the establishment of a national health management system should focus on family health advocacy.
Regarding the fact that health service utilization does not have mediating effects in the relationship between disease prevention and health outcomes, the root cause is that disease prevention literacy did not affect health outcomes. This is because the floating population is mostly young and has poor health-risk awareness. Meanwhile, disease prevention literacy involves professional preventive medical knowledge. Due to educational limitations, it is difficult for the floating population to have such understanding. In the future, subjects can be stratified by education level to examine the effects of disease prevention literacy on health outcomes. Given the current COVID-19 pandemic, the mobility of the floating population poses a huge health risk. Hence, there is an urgent need to strengthen their health literacy regarding disease prevention to reduce the risk posed by COVID-19.
This study has some limitations. First, the health literacy scale used in this study is a typical scale for critical health literacy, but there is a lack of scales for functional health literacy and communicative health literacy. In the future, the All Aspects of Health Literacy Scale (AAHLS) can be used to measure health literacy in the floating population as it is suitable for evaluating functional, communicative, and critical health literacy. In addition, there is a risk of bias when a self-evaluated health literacy tool is used since widespread optimism/pessimism and memory deficiencies can affect the outcomes. Second, the cross-sectional design did not allow for causality deduction. Conducting longitudinal studies and reassessing health outcomes will help identify causality in health literacy. Moreover, the samples came from three prefecture-level

CONCLUSION
This study found that health service utilization had partial mediating effects in the effects of health literacy on health outcomes. Health literacy affects the proactiveness of the floating population in health service utilization through healthcare literacy and health promotion literacy, thereby affecting health outcomes. Improving health literacy in the floating population will help improve their health outcomes. At the same time, health service providers need to enhance the diversity of health services and ensure that the floating population has the necessary external conditions to improve their individual health.