Analysis of Smart Home use based on the Degree of Health-related Risk Variation: A Cross - Sectional National Survey in China

Digital health has become a heated topic today and smart homes have received much attention as an important area of digital health. However, most of the existing studies have focused on discussing the impact of smart homes on people or the attitudes of older people towards smart homes. Only few studies have focused on relationship between health-related risks and use of smart homes. Aims To investigate the association between health-related risks and the use of smart homes, provide new recommendations to promote the implementation of digital health strategies and achieve health for all. Methods We used data from 11,031 participants aged 18 and above. The population was clustered based on ve health-related risk factors: perceived social support, family health, health literacy, media use, and chronic diseases self-behavioral management. A total of 23 smart homes were categorized into three sub-categories: entertainment smart home, functional smart home, and health smart home. We analyzed demographic characteristics and utilization rate of smart home across different cluster.


Introduction
Along with accelerated industrialization, urbanization, and population aging, China's disease spectrum continues to change. The death rate from chronic non-communicable diseases is in the proportion to 88% of all deaths. The resulting disease burden accounts for over 70% of the total disease burden. China government developed a "Healthy China" strategy in 2017 to improve the health literacy of residents, prevent diseases, and improve the quality of life of residents [1]. The Internet of Things has experienced rapid growth in the past decade, covering many elds, critical digital health. Following the development trend of technology, the World Health Organization has proposed a global digital health strategy for 2020-2025, advocating the promotion of digital health and the application of digital health to achieve the goal of universal health [2]. To help drive the development of health in China, the government is also using Jachan DE compared older people's satisfaction with smart homes with traditional mobility support tools and value for money. Overall, users rated all installed smart homes higher in terms of satisfaction. In terms of value for money, smart homes are of higher quality but also more expensive. The authors suggest that products can be modularized, thus reducing smart home product prices [18].
Different groups of people have various needs for smart homes, and expanding the smart home market requires understanding the needs of the audience on different sides of the product. The development of products cannot be separated from the support of technology and policies. In the context of the continuous development of the Internet and the much attention paid to the digital health strategy, understanding the differences in smart home needs of people with different health-related risks is conducive to smart homes producing products that meet consumer needs and promoting the development of this market, which can also, in turn, promote the effective implementation of the digital health strategy.

The present study
The role of health-related risks in the link between smart home use has not been explored in previous studies, despite the profound impact of health-related risks on the smart home use. The aim of this study is to examine the role of health-related risk in linking smart home use in China. Based on the above literature review, the following hypotheses are proposed.

Hypothesis 1
The health-related risk is negatively associated with smart home use. The lower the health-related risk is, the better the smart homes are used.

Hypothesis 2
Different health-related risk groups will prioritize different types of smart homes.

Data and Procedure
The data used in this study is conducted in 23 provinces, 5 autonomous regions, and 4 municipalities directly under the central government from July to September 2021. The survey is a multi-stage sampling, using the random number table method to select 2-6 cities from each noncapital prefecture-level administrative region of each province and an autonomous region, a total of 120 cities; based on the data results of "the seventh national census in 2021", quota sampling (quota attributes are gender, age, and urban-rural distribution) was conducted for 120 urban residents, so that the gender, age and urban-rural distribution of the samples basically conform to the demographic characteristics. Finally, 11031 valid questionnaires were obtained that have high quality and accurate national representation and comply with the ethical review rules (JNUKY-2021-018).

Variables Characteristic Variable
The characteristic variable in this study included respondents' socio-economic background (age, gender, income, Hukou, residence, education, public insurance, location recently, chronic disease, Disability, work status and Politics), family characteristics (Marriage, family type, number of children, household) and lifestyle (drinking status). See Supplementary Table S1 for details of de nitions and classi cations. Smart Home Use This paper investigates the Smart Home Use (SHU) of people for 23 kinds of smart homes, and divides them into three categories according to the functions of smart homes: entertainment SHU (consists of Smart TV, VR glasses, body sensing car, smart speaker), function assistance SHU (consists of a smart robot, smart lighting, smart washing machine, smart switch, smart door lock, smart toilet, smart mosquito repellent, electric curtain, smart air conditioner, smart clothes hanger, smart monitoring) and health SHU (consists of sports bracelet, temperature and humidity sensor, smart socket, danger button, smoke transducer, body fat scale, air puri er, smart medicine cabinet ). At the same time, the utilization rate of smart homes and the composition ratio of people are analyzed.

Health-related Risk
We identi ed ve factors that in uence health-related risk based on previous studies: perceived social support [19], family health [20], health literacy [21], media use [22], and chronic disease self-behavioral management [23]. These ve aspects were negatively associated with health-related risk, and we classi ed health-related risks according to the distribution characteristics of these ve aspects.
Perceived social support was measured using The Perceived Social Support Scale (PSSS) based on the Zimet Perceived Social Support Scale. A 12-item scale divided into three dimensions: family support, friend support, and other supports, as shown in Table S2 in the attached table. "strongly disagree", "slightly disagree", "neutral", "slightly agree", "agree", "Strongly agree" seven options, these seven options are assigned a score of 1-7 (Strongly disagree = 1). The higher the score, the higher the perceived social support. The alpha coe cients for family support, friend support, other supports and the full scale were 0.87, 0.85, 0.91 and 0.88 for the sample of 275 cases (139 males and 136 females) respectively, with retest reliability of 0.85, 0.75, 0.72 and 0.85 [24].
Family health was measured using the Family Health Scale-Short Form (FHS-SF) based on the AliceAnn Crandall [25] (Supplementary Table S3). For each item, subjects rated "strongly disagree", "somewhat disagree", "neither agree nor disagree", "somewhat agree " and "strongly agree", with the three dimensions (7 items) other than "family health resources" being assigned a value of 1-5 in order ("strongly disagree" = 1), while the three items for "family health resources" were assigned the opposite value ("strongly agree"=1). Cronbach's alpha for the 10-item scale was 0.80 and Cronbach's alpha for the FHS-SF was 0.84.
Health literacy was measured using the Short-Form Health Literacy Instrument (HLS-SF12) developed by Tuyen V Duong [26]. The scale has 12 items covering the three health domains of health care, disease prevention, and a health promotion, as detailed in Table S6 in the Supplementary Material. "very di cult", "di cult", " easy", and "very easy" for each item, in order of assignment from 1 to 4 ("very di cult" = 1). The higher the score is, the higher the health literacy is. This scale has high reliability with a Cronbach's alpha of 0.85.
In order to nd out how the participants use the media, we developed our own media use scale with seven items. The scale has ve options: "never use", "occasionally use", "sometimes use", "often use" and "almost every day" that are assigned a value of 0-4 in order (never use =0), see Table S4 in the Supplementary Material. Chronic disease self-behavioral management was measured using the Chronic Disease Self-management Study Measure (CDSMS) developed by Lorig [27,28]. The scale is divided into two sub scales: self -management behavior and self-management effectiveness. The self-management behavior scale consists of 15 items including exercise, cognitive symptom management practices, and communication with a doctor. The ve items were rated on a scale of 0-4 (not done=0), with higher scores resulting in higher status. See Table S5 in the Supplementary Material for details.
The Cronbach coe cient for the CDSMS was 0.72-0.75.

Statistical Analysis
For segmentation, non-hierarchical K-means cluster analysis was conducted based on 5 health-related risk-related factors: family health, health literacy, chronic self-behavior management, perceived social support, and media use.
Suppose a factor has a relatively large cluster center value. In that case, it can be characterized as a cluster that is positively affected by the factor. Then we conducted a t-test to explore the differences in the health-related risk-related factors, according to the clustering.
Firstly, we made a descriptive analysis of the demographic characteristics and the characteristics of each Characteristic Variable in the three clusters. The number distribution of each variable in the three clusters, the composition ratio distribution of each variable classi cation in the same cluster, and the proportional distribution of the same variable classi cation in the three clusters is counted. Chi-square veri cation was performed by comparing these data. Thirdly, we also make a descriptive analysis on the overall use of smart homes based on demographic characteristics. The smart home is divided into three categories, plus the overall unused category, and the number of users, the composition ratio, and the utilization rate of the following variables in these four categories are counted. Then, perform chi-square veri cation. Fourth, we made a descriptive analysis of the needs of the three groups for the speci c 23 smart homes under the three categories. We counted the number and the composition ratio of the three groups. Then, perform chisquare veri cation. Fifthly, according to the above analysis, we mainly focus on the differences in the demand for smart homes among the three variables of residence, gender, and age, and then make a descriptive analysis. The number and the composition ratio of 23 types of smart homes used by people of three age groups with urban and rural residence, male and female gender are counted. Then, perform chi-square veri cation.

Results
Segmentation Based on Health-related risk As the clustering was more balanced in each group when clustered into three groups and there were signi cant differences in health-related risk characteristics among the groups, the clustering can be judged that is convincing when clustered into three groups. Table 1 showed the results of the clustering de ned into three groups. The analysis shown that there was a signi cant difference among three clusters in Chronic self-behavior management, Health literacy, Media use, Perceive social support and Family health (p< 0.05). The clusters were de ned and named based on the level of health-related risks and characteristics. cluster 1 'low risk' is a group that the mean values of each variable were each higher than their overall respective mean values. they take the lowest health-related risk. cluster 2 'middle risk' is a group whose mean values are above average except for media use and chronic self-behavior management. their health-related risks are moderate but still require control. cluster'3 'high risk' is the highest health-related risk group that means values lower than average. the difference among the factors for each group can be seen in gure 1. Table 2 shows the demographic and covariant characteristics of each cluster. There was a signi cant difference in some Characteristic Variables among the three groups. The proportion of females in Cluster1, Cluster2, and Cluster 3 was higher. Among Cluster1, Cluster2, and Cluster 3, the proportion of young and middle-aged people aged 19 ~ 45 was higher, and the proportion of people aged over 91 was lower. The Middle Income Group of Cluster 1 with income from 3001 to 7500 and the High Income Group of above 7501 were higher than the Middle Income Group of Cluster 2 and Cluster 3 with income from 3001 to 7500. Cluster 1, Cluster 2, and Cluster 3 have a higher proportion of the urban population, and the difference is more obvious in Cluster 1. The other parties in Cluster 1, Cluster 2, and Cluster 3 had higher percentages, and the disparity was even greater in Cluster2. The proportion of postgraduates and doctorates in Cluster1, Cluster2, and Cluster 3 was higher than that in middle school and Lower Education, and the disparity in Cluster 1 was more obvious. The proportion of in-services in Cluster1, Cluster2, and Cluster 3 was higher than that of retired.

Descriptive statistics and correlations
Cluster 1, Cluster 2, and Cluster 3 had a higher proportion of Nuclear family types and a lower proportion of singleparent families. The proportion of the urban population in Cluster 1, Cluster2, and Cluster 3 was higher. The proportion of married and unmarried people was higher in Cluster 1 and Cluster3, and the proportion of married people was higher in Cluster 2. The proportion of childless people in Cluster 1, Cluster2, and Cluster 3 was higher, and the difference was even greater in Cluster 1. A higher percentage of cluster 1, Cluster 2 and Cluster 3 had one or two people living with them. The proportion of the population using public health insurance is higher in Cluster 1, Cluster 2 and Cluster 3.
There were no chronic diseases in Cluster 1, Cluster2, and Cluster 3, but a high percentage of people with disabilities.
Among Cluster 1, Cluster 2 and Cluster 3, the proportion of people who had recently consumed alcohol was higher. Table3 describes the differences in the demographics and Characteristic Variables for smart home usage. The gender factor in the overall use has the remarkable difference, displays for the female overall use rate to be higher than the male. There were signi cant differences in age factors. The overall rate of using entertainment smart home was higher among 19-45-year-olds, and the rate of using smart home for functional and health was higher among 91-year-olds. The income factor has the remarkable difference, manifests for the income above 7500 high income crowd each kind of smart home use ratio to be high. There are signi cant differences in household factors, and the proportion of various kinds of smart home in rural household is high. There are signi cant differences in the political landscape, with a higher proportion of CPC smart homes being used. The educational factor has the remarkable difference, displays for the master degree above crowd to use each kind of intelligent home the proportion to be big. There were signi cant differences in the working environment, with a higher proportion of in-services using various types of smart home. There were signi cant differences in household type factors, with people with Nuclear family type using a higher proportion of different types of smart home. There is a signi cant difference between recreation and health in the factors of recent residence, which shows that the proportion of recreation is higher in the western region and health is higher in the Eastern Region. There are signi cant differences in the factors of residence, indicating that the proportion of urban population using various kinds of smart home will be higher. There were signi cant differences in factors related to marital status, which indicated that unmarried people had a higher overall utilization rate and a higher proportion of healthy smart homes, while divorced people used a higher proportion of functional smart homes. There was a signi cant difference in the number of children, as demonstrated by the higher overall utilization rate of one child and the higher utilization rate of recreational smart homes, and the higher proportion of functional and health analogs among the childless population. There are signi cant differences in the factors of public insurance, as a result of the high usage rate of various smart homes among the insured population. There were signi cant differences in the factors of chronic diseases, and the proportion of using various smart home without chronic diseases was higher. There was a signi cant difference in the factors of drinking, which showed that the proportion of using health smart homes was higher among the people who drank before 30 days. See Table S7 in the Supplementary Material for more details on demographic differences in overall smart home use. The analysis of demographic differences in users is detailed in Table S7 in Supplementary Material.  Table 4 showed that there were signi cant differences among the three groups in the needs of 23 smart homes in three categories. Among them, the low risk group had higher demand for smart homes, especially functional homes, with a utilization rate of 77.42%. It was much higher than that in the middle risk groups and high risk groups. Except that the demand for entertainment smart homes in the medium-risk groups was higher than that in the high risk group, the demand for functional and health smart home was lower than that in the high risk group. For the entertainment class, smart TV had the highest utilization rate, while VR glasses and body sensing cars had a lower utilization rate. For the functional class, the utilization rates of smart washing machines and smart air conditioners were high, while the utilization rates of electric current, smart clothes changers, smart mosquito reply, and smart robots were low. For healthsmart home, sports brace, body fat scale, and air puri er were used more frequently, while temperature and humidity sensor, danger button, smoke transmitter, and smart medicine cabinet were used less frequently. Subgroup analysis According to Table 5, there were signi cant differences in the demand for smart homes between urban and rural populations. The urban population accounts for more than 70%, while the rural population accounts for less than 30%. This difference was caused by the differences in urban and rural economy, urban and rural education, the actual use place of smart home, and the practicability of the smart home. There were signi cant differences in the demand for smart homes among people of different ages. The middle-aged (45 ~ 49 years old) account for more than 50%, while the elderly (76 ~ years old) were generally less than 10%. This difference was caused by educational level, cognitive ability, physical action, and economic status There were gender differences in the needs of VR glasses, body-sensing cars, smart TV, smart brace, smart medicine cabinet, smart socket, temperature, and humidity sensors, and the demand of men was higher than that of women. This difference was caused by physiological differences, family structure, and income differences. At the same time, gender differences also had differences in the demand for smart clothes changers, which shows that the demand of women was higher than that of men. The reason for this difference is that women pursue fashion and appearance more than men do. And women's clothing is more expensive than men's, so women will pay attention to the protection and proper custody of clothing.

Discussion
In the context of the COVID-19 pandemic, the importance and convenience of digital health are becoming more and more prominent [29,30], with telemedicine enabling people to have medical consultations at home and avoid infections, and many countries and regions are actively promoting the development of digital health [31,32]. Identifying the needs of different groups of people for smart homes is important to promote the implementation of digital health strategies, and we conducted a cluster analysis of the population based on health-related risks, divided into three groups: the low risk group, the medium-risk group, and the high risk group, and con rmed the differences in the needs of different groups of people for smart homes through research and analysis.
The usage rate of smart homes for the low health-related risk group was 86.97%, the usage rate for the medium healthrelated risk group was 79.23% and the usage rate for the high health-related risk group was 77.36%, the lower the health-related risk the better the usage of the smart home. The low health-related risk group had higher health literacy and chronic disease self-behavior management had a stronger health mindset, were better able to self-manage their diseases [33,34], and were more likely to use smart homes for health management. In addition to this, the population of the low health-related risks group had the best media use among those, with greater information exposure, and more likely to learn about smart home-related information. In addition to these factors, sociodemographic characteristics also had an impact on smart home use in the three groups The age distribution of the low risk group was younger than the other two groups, with younger people using smart homes better than older people, and Alhuwail D's study also indicated that younger people used smart devices more than older people [35]. Our analysis suggests that age affects smart home use in four main ways. Firstly, consumer perceptions are different; while older people show positive attitudes during the experience of the entertainment smart home -VR glasses -they do not have a strong desire to buy them, believing that smart homes are unnecessary in their lives, while younger people associate the experience more with the content being fun [36]. Secondly, the prevalence of chronic diseases is higher among the elderly [37][38][39], and studies have shown that three-quarters of the elderly in China suffer from at least one chronic disease [40], and health a smart homes can help them with daily check-ups of diseases and daily behavior management [41][42][43]. The China Quality of Life Development Report for the Elderly (2019) shows that about 29.6% of the elderly in China have not attended school, 41.5% have a primary school education, and 25.8% have middle and high school education. It can be seen that the literacy level of older people in China is relatively low [44], and the low level of education limits the use of smart products by older people. Finally, due to the deterioration of physical functions, the elderly have di culty in understanding the operation of smart homes and are worried about not being able to use them independently without help [45,46].
The low risk group has the highest proportion of high-income people, and income is one of the in uencing factors for smart home usage. Smart home products have higher technical requirements for research and development, require a higher level of R&D talents, and have high manufacturing costs. Small companies also nd it di cult to enter this industry due to technical and nancial problems, making it di cult to achieve scale effects, which leads to the problem of high prices of smart home products, and many users also indicated during the survey that purchase cost is one of the main considerations [47].
The low risk group had a larger urban population, and place of residence was also a signi cant in uencing factor on smart home use, with urban use better than rural. According to the survey, 84.2% of urban students in the Washington State school district reported being able to use reliable broadband to watch instructional videos, while only 67.5% of rural areas agreed [48]. In China, the Internet gap between urban and rural areas is gradually decreasing, but there is still a 24.1% difference in penetration rates [49]. Smart homes are built on the internet, rural Internet infrastructure is weaker than urban areas, and the digital divide is a barrier to the spread of digital health in rural areas [50,51], which is one of the reasons for the low usage of smart homes in rural areas. In addition to this, there are differences in access to health information between rural and urban residents, with rural residents having less access to health related information from sources such as specialists, magazines and less frequent use of search engines than urban residents [52], which may contribute to rural residents not knowing enough about digital health and smart homes.
Although there are many differences in the demand for smart homes among different health-related risk groups, they also show similarities: men have a signi cantly higher purchase rate for VR glasses, body cars, smart bracelets, smart pillboxes, smart sockets, and temperature and humidity sensors than women, but women have a signi cantly higher usage rate for smart drying racks than men. This is mainly determined by lifestyle, income, and subjective willingness.
Currently, men generally have higher incomes than the income of women [53,54], which determines that men have stronger purchasing power for higher-priced smart homes. Secondly, men and women have different household lifestyles; women take on more household chores [55,56] and maybe more interested in smart drying racks related to household chores, while men have more leisure time at home and purchase more entertainment products. In addition, research has shown signi cant differences in the technology acceptance between men and women, with men showing a stronger intention to use technology [57,58], while women are more interested in fashion products during consumption [59] and maybe more interested in daily protection of clothing closely related to fashion.
Looking at the three broad smart home categories, smart TVs have the highest usage rate in the category of entertainment, with VR glasses and body sense cars having lower usage rates. This is related to the high frequency of use and usefulness of TVs as traditional homes [60], while VR glasses and body cars may show lower usage because our sample includes middle aged and elderly people, who are limited by their age and physical mobility [61] to use such exercise products. In the category of functional, smart washing machines and smart air conditioners were used more frequently, while electric curtains, smart hangers, smart mosquito repellents and smart robots were used less frequently. This is related to the in uence of the early use of washing machines and air conditioners that are widely used and highly practical, while electric curtains, smart drying racks and mosquito extinguishers are less practical and costeffective, and smart robots are narrowly used, expensive and technically immature, with most of the experiencers expressing a neutral attitude [62].In the health category, the usage rate of sports bracelets, body fat scales and air puri ers were high, while the usage rate of temperature and humidity sensors, hazard buttons, smoke sensors and smart medicine cabinets was low. This is related to the development of supporting facilities such as the Internet, the popularity of sports bracelets, body fat scales and air puri ers, and their low prices. The rest of the products are not as popular, cost-effective and practical, and have fewer manufacturers. We could not nd any information about these products on the o cial websites of the larger Chinese smart home manufacturers, Xiaomi and Huawei.
Implementing a digital health strategy requires a concerted effort by product manufacturers and the government. As older people consume products with greater consideration for ease of use and practicality, manufacturers should pay more attention to the older market and simplify the steps and the interface design of their products. Studies have shown that user involvement in the product design can effectively improve the quality, relevance and prevalence of work [63,64], and companies can recruit volunteers to participate in the design process as appropriate to the actual situation. Health products are of importance for aging [42], but the current smart home market has a small range of such products, so manufacturers need to implement the concept of digital health and increase the development, production and promotion of such products. Because of the poor usage of smart homes among the elderly, low-income groups, rural residents and women, manufacturers should pay more attention to the positive and negative factors in uencing their purchases and broaden the market for their products.
The high production costs of companies lead to the high pricing of products, which is one of the major obstacles to people using smart homes. The government can enhance policy guidance to attract investors and capital into the smart home market and reduce the production pressure on enterprises. In addition, the government should also encourage enterprises to invest in technology research and development to break through existing technical di culties, thereby reducing the di culty of product production and achieving lower product prices. Secondly, because there is also no uni ed industry-standard speci cation for smart homes in China [3], while different companies have different product compatibility [65], leading to the problem that elderly users feel more di cult to use in the process. The relevant authorities should formulate industry standards to address this issue and solve the problem of different product compatibility. Finally, the government also needs to pay attention to strengthening the construction of residents' health knowledge system, raising their health awareness, reducing their health-related risks, improving their ability and awareness of using technology for health management, and fully implementing digital health strategies to improve the quality of their healthy lives and achieve health for all.
Our study clustered the ve in uencing factors of health-related risk: family health, perceived social support, health literacy, chronic disease behavior management and media exposure into three groups: the low risk group, the mediumrisk group, and the high risk group, which contributed to a new way of segmentation. In addition to this, the relationship between health-related risk and smart home use has never been looked at in previous studies, but our study demonstrates that smart home use differs between health-related risk groups, with the low health-related risk groups showing better overall smart home use, and con rms that people currently prefer functional smart home products and that low health-related risk groups are also concerned about health products. This suggests that our research hypothesis that different levels of health-related risk in uence smart home use are valid. These ndings provide important new insights that have implications for the development of the smart home industry and the implementation of digital health strategies.
However, due to limitations in research experience and resources, our study also has certain limitations. Our survey was conducted through respondents retrospectively completing a questionnaire, which may be subject to recall bias.
Secondly, our study was a cross-sectional survey and was unable to demonstrate a causal relationship between healthrelated risks and smart home use, which could be explored in future studies through longitudinal research. Finally, we analyzed health-related risk through ve dimensions: family health, perceived social support, health literacy, chronic disease behavior management, and media exposure, possibly ignoring the effect of other variables on health-related risk.

HLS: Health Literacy Scale
This work was supported by the Medical Scienti c Research Foundation of Guangdong Province (A2020420), the 13th Five-Year Plan of Guangdong Province for Philosophy and Social Sciences (GD20XGL42) and the Guangzhou Science and Technology Plan Project (202103000037). The funders had no role in the study design, data collection, data analysis, data interpretation and report writing. The corresponding author has full access to all data in the study and is ultimately responsible for the decision to submit for publication.