Research on inuence of physical activity on medical expenditure in the elderly in China—Based on CHARLS in 2015

Becoming more and more serious of such problems as acceleration of aging and high medical burden of the elderly in China, research on the inuence of physical activity on medical expenditure in the elderly has attracted a lot of attention. In this work, Tobit model is established based on the data from China Health and Retirement Longitudinal Study in 2015 and Behavioral Model of Health Services Use theoretical framework. Compared with the inactive elderly, the medical expenditure of the elderly who participate in the level 4, 3, 2 physical activity decreasing step by step. After involving need factors in the model, the inuence of physical activity on medical expenditure varies from signicant to insignicant. The medical expenditure of the elderly who meet the physical activity amount is negatively correlated with the physical activity amount, and is lower than that of the elderly who don’t meet physical activity amount. Physical activity will not directly affect medical expenditure, but will affect medical expenditure through chronic diseases and self-rated health status. Our results provide theoretical analysis and empirical test for enriching physical activity guidelines and Behavioral Model of Health Services Use in the elderly in China.


Introduction
In recent years, the aging speed in China continues to accelerate. According to the latest data released in January 2020 by the National Bureau of statistics, by the end of 2019, the population aged 60 and above has reached 253.88 million, accounting for 18.1% of the total population, far exceeding 10% of the social standard line for aging. The aging population has brought a serious burden to our society. According to the latest data of the National Bureau of statistics, the total national medical expenditure in 2018 was about 5.9 trillion yuan, increasing by 12.4% compared with that in 2017, and the per capita health expenditure was 4237.0 yuan, increasing by 12.0% compared with that in 2017 1-2 . The factors in uencing medical expenditure are complex, and physical activity is one of the changeable factors. In the eld of sport science research, Yang Guang et al. 3 conducted a factor analysis on the in uence of physical activity on medical expenses in the elderly in Hegu area of Sendai Japan, through questionnaire survey; Li Wenchuan 4 conducted an empirical study on the reduction of medical expenditure by physical exercise of the elderly in Shanghai.
Lehnert t et al. 5 pointed out that the utilization of medical services is mainly related to physical factors in health conditions, rather than psychological factors. Physical factors mainly refer to chronic disease conditions, physical function conditions and physical activities. Guo Jialiang et al. 2 believe that among the many factors in uencing the health and health expenditure of the elderly, physical activity is also becoming one of the main in uencing factors. Compared with other factors, physical activity is the most promising factor to be improved through action 6 . Based on the cross-sectional data of China Health and Retirement Longitudinal Study (CHARLS) in 2015, this paper uses the Behavioral Model of Health Services Use (BMHSU) and Tobit model to study the relationship between physical activity and medical expenditure of the elderly, and try to analyze the relationship between the frequency, intensity, duration and form of physical activity and medical expenditure of the elderly.

Methods
Study design. When studying the in uencing factors of personal medical expenditure, health status is regarded as the most direct in uencing factor. Hu Hongwei et al. believe that chronic diseases and self-rated health signi cantly affect the medical expenditure of the elderly 7,8 . In addition to health status factors, Liu Guoen et al. believe that the income, medical insurance and pension of the elderly will also in uence their medical expenditure 9-11 . In addition, education level, marital status, age, gender and urban and rural identity will also in uence medical expenditure 7,11 . According to the research on the impact of physical activity on medical expenditure of the elderly, scienti c physical activity has signi cant effect on reducing the incidence of diseases, reducing medical demand and saving medical expenditure 12 . According to the intensity, amount and duration of exercise, Yang Guang 3 divided the amount of physical activity of the elderly over 70 years old in the Hegu area of Sendai Japan into three levels. With the decrease of intensity and amount of exercise, there are signi cant differences in medical expenditure among the three groups.
In summary, the factors that affect the personal medical expenditure of the elderly in China are relatively complex, involving health status, economic conditions, social security, socio-demographic characteristics, physical activities and many other aspects. Professor Andersen's BMHSU provides a comprehensive and systematic theoretical framework for the research in this eld. In this dynamic and interactive relationship model (Figure 1), demographic characteristics affect medical expenditure, and personal health practices affect demographic characteristics. Demographic characteristics include three aspects: need factors, enabling resources and redisposing characteristics 13 . In addition to the medical service utilization behavior, the health behavior in the theoretical model also includes personal health practices, such as diet and exercise 14  Selection of variables. The per capita medical expenditure of the elderly is the dependent variable of this study, which is the sum of the monthly outpatient expenditure and the annual hospitalization expenditure in the questionnaire 7,17 , including the reimbursement of medical insurance and the part of individual selfpayment.
According to BMHSU, fteen variables including need factors, enabling resources, predisposing characteristics, personal health practices and the use of medical services are selected as independent variable and control variable (Table 1), of which personal health practices mainly select sports dimension, including four levels of physical activity. Physical activity refers to any physical activity that results in energy consumption caused by skeletal muscle contraction, which is composed of four types of physical activities: occupational, tra c, housework and leisure 18,19 . Intensity, duration weekly and duration each time are used as index to measure physical activity 20 . The impact of physical activity on health depends on the way, intensity, duration, frequency and total amount of physical activity 21 .
In this study, the selection of independent variable --four levels of physical activity is derived from the question of "physical activity" in the 2015 CHARLS questionnaire: "How long do you usually take in these activities (including high intensity, medium intensity and walking) in a week?". According to WHO physical activity guidelines for the elderly 25 , Chinese adult physical activity guidelines of at least 30 minutes per day, Chinese Ministry of Health's proposal of "ten thousand steps per day", and the related research on physical activity amount, the independent variables are divided into four levels 3 , and the level 4 physical activity group includes: high intensity exercise (more than 75 minutes per week, no less than 10 minutes each time), moderate intensity exercise (more than 150 minutes per week, no less than 10 minutes each time), and walking (210 minutes a week, no less than 10 minutes each time); The level 3 physical activity group includes: no high intensity exercise, moderate intensity exercise (more than 150 minutes per week, not less than 10 minutes each time), walking (210 minutes a week, not less than 10 minutes each time); The level 2 physical activity group includes: no high and moderate intensity exercise, walking (210 minutes a week, not less than 10 minutes each time); The level 1 physical activity group includes: the elderly who doesn't meet the above three levels of intensity. The level 1 physical activity group doesn't meet the physical activity, the other three groups meet the physical activity. Ethics approval and consent to participate. For all experiments on humans, we con rm that all methods were carried out in accordance with relevant guidelines and regulations. The CHARLS received ethics approval from both a national government administration and an institutional review board or ethics committee. Only adolescents who provided written/verbal consent participated. As the current study used retrospective publicly available data, we did not require ethics approval from any Institutional Ethics Review Committee for this secondary analysis.
The CHALRS questionnaire refers to international experience, including the American Health and Retirement Survey (HRS), the British Elderly Tracking Survey (ELSA), and the European Health, Old Age and Retirement Survey (SHARE), etc. Multi-stage sampling is adopted in the project, and PPS sampling method is adopted in both county/district and village sampling stages. The response rate and the quality of CHARLS data are among the best in similar projects in the world, and the data has been widely used and recognized in academia.

Comparison of physical activities of the elderly with different characteristics
The data shows that there are signi cant differences in medical expenditure, gender, education level, marital status, pension, self-rated health status, and whether there is disability or not ( Table 2).
The medical expenditure and physical activity of the elderly who meet the physical activity shows a negative correlation trend; the proportion of men in level 4 physical activity group is higher than that in the other three groups; the education level of the elderly who meet the physical activity amount is higher than that in the elderly with insu cient physical activity; the self-rated health status of the elderly who meet the physical activity amount is more positive than that in the elderly who doesn't meet the physical activity.
For the elderly who don't meet the physical activity amount, the majority of them are women, whose age are higher than the average age, which is 70.52 years old. Their education levels are mainly below primary school, and cohabiting with spouse are also below the average level. The proportion of buying medical insurance and insurance is relatively low, and the personal income is lower than the average level. The selfrated health status is relatively negative, and the proportion of disabled people is more than a half. According to the physical activity of the disabled elderly, 80.41% of them meet the physical activity, and the number of persons participating in the second, third and fourth level physical activity is distributed more evenly, which indicates that most of the disabled elderly have certain demand for physical activity. For the elderly patients with chronic diseases, 84.48% of the elderly meet the physical activity level, and the number of persons participating in the second, third and fourth level of physical activity is distributed more evenly. Among the four groups of elderly physical activity, there are no statistically signi cant differences in the comparison of physical intensity in the ve indicators of age, urban and rural household registration, personal income, medical insurance and whether there is chronic disease.

Results of Tobit model
According to model 1, physical activity of the elderly will signi cantly in uence the medical expenditure of the elderly in China. The lower the physical activity, the higher the medical expenditure of the elderly (Table   3)

Discussion
In the whole life course, physical activity is closely related to health. A large number of cross-sectional studies and longitudinal studies have shown that the relative risk of limited activity function is reduced by 50% among people who regularly participate in at least moderate intensity of physical activity. For the elderly, physical activity can reduce anxiety and depression and improve self-con dence by maintaining muscle strength and cognitive ability. Randomized controlled trials show that physical activity can increase social participation, maintain social networks and intergenerational connectivity. In addition, a large number of literature shows that physical activity has an impact on health status, but this study concluded that due to the impact of health, the in uence of physical activity on medical expenditure is signi cant or insigni cant.
In Tobit model 4, after adding the need factors, there are statistical effects on in uence of health and whether suffering from chronic diseases on the elderly medical expenditure, which also proves the direct impact of health on medical expenditure. The self-rated health status of the elderly will signi cantly in uence the medical expenditure of the elderly in China. The more negative the self-rated health status is, the higher the medical expenditure is. Maybe when the elderly self-rated health status is poor, outpatient treatment is frequent 26 . In addition, disease is the most direct in uencing factor. The prevalence of chronic diseases in the elderly in China is relatively high, with an average of 1.68 chronic diseases, of which 73.21% suffer from one or more chronic diseases. Elderly patients with chronic diseases have a long and large demand for medical care. In addition, once the chronic diseases recur, the treatment cost is very expensive 23 . It also further shows that the average annual medical expenditure of the elderly patients with chronic diseases is higher than that of the healthy elderly. Whether suffering from disabilities has no signi cant impact on the medical expenditure of the elderly, which is the same as Ma Aixia's analysis. The reason may be that most disabilities, such as physical disability, deafness and blindness, do not need continuous medication or hospitalization, so the impact on medical expenditure is limited 6 .
The study also nds that the higher the education level is, the higher the medical expenditure is. The medical expenditure of the elderly with diploma of primary school, middle school, high school and above are 1.01, 1.34 and 1.77 times of that of illiterate people respectively and higher education level increases the ability of the elderly to use medical services 27 , which is the same with Liu Guoen et al.'s study result that income has a signi cant impact on medical expenditure, and the medical expenditure of high-income population is signi cantly higher than that of the low-income population 10 . The higher the income level of the elderly is, the more attention they pay to their health status, so the medical expenditure is more 23 .
According to the results of one-way ANOVA of physical activity of the elderly with different characteristics, the medical expenditure and physical activity of the elderly who meet the amount of physical activity show a negative correlation. At the same time, Tobit model 1 shows that physical activity has a signi cant impact on medical expenditure. The medical expenditure of elderly people who participate in high-intensity or moderate intensity exercise and walking exercise is far lower than those who only participate in walking exercise and who walk for 210 minutes or more a week, that is there is a difference medical expenditure between the elderly people who participate in walking only and those who participate in middle and high intensity of used a retrospective cohort study and found that the average annual medical expenditure of the elderly over 65 years old who participated in community exercise programs in the past three years was far lower than that of those who did not regularly participate in physical exercise. However, there is a lack of research on the difference of medical expenditure between walking and middle or high-intensity exercise.
In addition, because predisposing characteristics will not directly in uence personal medical expenditure, but indirectly play a role through in uencing enabling resources and need factors 8 , as the basis in building model, the predisposing characteristics are analyzed rstly and included the direct in uencing factors-need factors in this paper. In this study, because the occurrence of diseases or the decline in health status are the prerequisites and direct in uencing factors for individuals to use medical services 8 , therefore, with the addition of need factors in model 4, the impact of physical activity changes is signi cant to insigni cant. To a certain extent, physical activity will not directly in uence medical expenditure, but may in uence that by chronic diseases and self-rated health status, which con rms the exploration of the dynamic relationship between health behavior and demographic characteristics by BMHSU model from the side 14 . Through many quantitative empirical studies, scholars at home and abroad have shown that there is a certain correlation between physical activity and medical expenditure of the elderly, but they have not explained the in uencing path at the same time. At present, some studies take health-related variables as mediation factors for path analysis. A public health survey in Sweden studied the mediation role of self-rated health between physical activity and medical service utilization through path analysis 30 . The in uence of physical activity on medical expenditure is a long-term effect, so it is suggested that the elderly should be encouraged to improve their need factors through physical activity, thus affecting medical expenditure. However, for its mechanism, it is suggested to use path analysis to explore, which can be further demonstrated in future research.

Declarations
Data availability The data that support the ndings of this study are available from National School of Development, Peking University. Restrictions apply to the availability of these data, which were used under license for this study. Data are available at http://charls.pku.edu.cn/pages/data/111/zh-cn.html with the permission of Peking University.

Competing interests
The authors declare no competing interests.