The sample for this study was derived from the 2020 China Health and Retirement Longitudinal Study (CHARLS) dataset. The CHARLS sample covers 150 counties and districts, 450 villages and communities, randomly selected from across the country, with more than 10,000 households and approximately 18,000 individuals aged 45 years and older. It is, therefore, a good representation of China’s middle-aged and older adult population (27). In this study, those aged 45 and above were selected as the research subjects, and samples with missing key variables and samples with a community sample size of less than 20 were excluded, resulting in 16,188 valid observations.

Health behavior is the core explanatory variable of this study. Five indicators related to healthy living habits were selected, namely smoking, drinking, napping, exercising, and sleep duration, as were five indicators related to health risk prevention behaviors, namely outpatient behaviors, consumption of health care products, health check-ups, social pension insurance, and social medical insurance, all taken from the China Health and Retirement Longitudinal Study. The definitions of the specific indicators are given in Table 1. The study drew on the average Euclidean distance method used by Nan (28) to measure the accessibility of public health services, and the health behavior sub-indicators were summed to obtain a health behavior index for middle-aged and older adult people. First, the indices of healthy living habits and health risk prevention were summed separately, and then the two indices were synthesized into a total index of health behaviors. This method satisfies the characteristics of the indicator units being independent, bounded, and monotonic, thus effectively measuring the equilibrium within the same group.

The explanatory variable in this study is the peer effect. Following the example of prior studies, this study set individuals living in the same community/village as belonging to the same peer group (29). The peer effect was calculated as the mean value of the level of health behaviors in the same community/village other than the respondent himself/herself. This study included three moderating variables: Internet use, Internet use proficiency, and Internet use preference. Internet use was measured by the item, “In the past month, did you go online?,” while Internet use proficiency was measured by the item, “Do you use a mobile phone to make payments, such as with Alipay, WeChat wallet, etc.?”; answers were coded as 1 = yes, 0 = no. Internet use preference was measured by the item, “What do you usually do online?,” with possible answers being chatting, watching the news, watching videos, playing games, and finance.

This study also controls for other variables that may influence an individual’s health behavior. These include age, age squared, gender, education level, community type, health level, household type, marital status, whether or not they are cohabiting, work status, and total annual income.

This study used Stata16 for statistical analysis of the data. The Ordinary Least Squares (OLS) model, Quantile Regression, and the Virtual Cohort Test (VCT) were used to analyze the influence of group effects on the health behaviors of middle-aged and older adult people. The interaction term model was used to test the moderating role of Internet use and Internet proficiency in the group effects of health behaviors on middle-aged and older adult people, and subgroup regression was used to test the heterogeneity of populations, communities, and Internet use preferences. The test level was α = 0.05.

A total of 16,188 middle-aged and older adult people were included in this study, of whom 7,696 (47.54%) were men and 8,492 (52.46%) were women. 7,479 (46.20%) were aged 45–59 years, and 8,709 (53.80%) were aged 60 years and above; 13,841 (85.50%) were with spouses, and 2,347 (14.50%) without; the community type was 10,733 (66.30%) rural and 5,455 (33.70%) urban; 12,440 (76.85%) lived in agricultural households, and 3,748 lived in non-agricultural households. With regard to educational level, 10,627 (65.65%) had completed primary school or below, 5,279 (32.61%) had completed secondary school, and 282 (1.74%) had attended tertiary education. With regard to working status, 11,045 (68.23%) were employed and 5,143 (31.77%) were either unemployed or retired.

Table 2 presents the estimation results from the Ordinary Least Squares (OLS) model and Quantile Regressions. The results of the OLS model show that the average level of health behavior of middle-aged and older adult people in the community other than the person himself had a significant effect on individual health behavior (β = 0.5341, p < 0.001). Specifically, for every 1-unit increase in the average health behavior level of the community, the health behavior of individual middle-aged and older adult people increased by 53.41 percentage points. Quantile regression was used to determine whether the peer effect for middle-aged and older adult people with different levels of health behaviors differed significantly. The estimated coefficients of the quantile regression can be interpreted as the marginal effects of the explanatory variables at specific quantile points; Table 2 presents the regression results at the three quantile points of 0.25, 0.5, and 0.75. The results show that the average level of health behavior of others in the community had a significant positive effect on individual health behavior, and that effect increases first and then decreases as the quantile points increase, with estimated coefficients of 0.4403, 0.4693, and 0.3985, in that order. This suggests that middle-aged and older adult people with average levels of health behaviors were more susceptible to the influence of the health behaviors of others, relative to the lower and higher groups.

This section concerns the moderating effects of Internet use and Internet proficiency on the peer effect of health behaviors on middle-aged and older adults and the related population heterogeneity. By adding the cross-multiplier terms of Internet use and peer effect, the Ordinary Least Squares model was used to regress the sample of middle-aged and older adults, the sample of rural and urban communities, and the total sample. The results are shown in Table 5.

This indicates that Internet use positively moderated the health behavior peer effect in the older adult population.

Table 6 presents the results of the moderating effects test and the population and community heterogeneity analysis for Internet proficiency.

Internet usage preferences were clustered into three groups based on five activities—chatting, watching news, watching videos, playing games, and managing money—measured by the question, “What do you usually do online?” These five activities were K-means clustered into three groups: cultural and entertainment (watching videos, watching news, playing games, and managing money), social and entertainment (chatting and watching videos), and comprehensive (chatting, watching videos, watching news, playing games, and managing money). Among the participants, there were 2,754 middle-aged and older adult people whose Internet use preferences were cultural and entertainment, accounting for 41.73% of the total number of people using the Internet; 1,239 people, or 18.77%, had a preference for social and entertainment; and 2,607 people, or 39.50%, were assigned to the comprehensive group. Table 7 presents the regression results for the peer effect of health behaviors on middle-aged and older adult people with different types of Internet use preferences. The results show that the size of the influence of the peer effect on those with different Internet use preferences was in the order of social-entertainment type (β = 0.6250, p < 0.01), cultural-entertainment type (β = 0.5835, p < 0.01), and comprehensive type (β = 0.4622, p < 0.01).

The average level of health behaviors of middle-aged and older adult people in the community was positively correlated with individual health behaviors, suggesting that there is a peer effect on the health behaviors of middle-aged and older adult people. This result is consistent with the findings of Xie Donghong et al. (15), who found that an improvement in the average level of health in the community promotes proactive health-seeking behaviors. According to the theory of the peer effect, the performance of specific behaviors by an individual is often influenced by the behaviors and characteristics of other individuals in their social group, leading to similarities with the group’s behavioral performance (30); health behaviors are one type of such behaviors that are influenced by the surrounding group. The community is the main place where residents conduct their daily activities and an important unit for the peer effect (31). The peer effect of health behaviors on middle-aged and older adult people may exert its effects through the following three mechanisms.

The first is the social interaction mechanism: social interactions may change an individual’s perceptions and ideas about health, and even influence their judgments about health behaviors. The older adult in a community can form a relational social network based on trust and reciprocity, and, in the process of social interaction, they will pass on and exchange healthcare information, thus guiding other individuals in making decisions regarding their health behaviors (32).

The second is a culture-driven mechanism: due to the collectivist mindset that is dominant in the older group, the middle-aged and older groups are more proactive in seeking out healthy behaviors and aligning themselves with group behaviors when there is an increase in the overall level of health behaviors in the community (33). This peer effect of health behaviors occurs through interactions between neighbors, and the stronger the interactions between middle-aged and older adult people in the community, the greater the influence on individuals. Traditional Chinese neighborhood culture emphasizes the idea that neighbors are good companions. The frequent interactions among neighbors in the community under the acquaintance society promote the flow of health information.

The third is the social norms mechanism: based on social rules, people subconsciously look to the dominant behavioral norms as a guide to their own behavior in a given social situation (34). The Outline of the “Healthy China 2023” Plan, under the theme of “building and sharing health for all,” makes the building of healthy communities an important part of the construction of a healthy China. Each community promotes healthy behaviors and discourages unhealthy behaviors by optimizing their social environment and formulating relevant systems such as the Healthy Community Rules and System Measures, thereby raising the average level of healthy behaviors in the middle-aged and older adult people in the community. The health disparity utility of communities and groups reinforces each other, and the community, as an important unit of peer effect, can be a reliable vehicle for health messaging and health policy implementation. Therefore, the government should give full play to the dynamic role of the community, encourage its guidance and support for community health promotion work, and promote the building of a healthy community.

In summary, there is a peer effect on the health behaviors of middle-aged and older adult people in a community in the context of Chinese collectivist culture. The results of this study are informative for the formulation of guidance for community residents’ health behaviors in collectivist-based social structures. In addition, the results of this study provide lessons relevant to the health level of immigrant communities in an individualism-based social structure. Immigrant communities are often culturally homogeneous and foster frequent communication and interaction, facilitating peer phenomena. Enhancing the health of others through the influence of group members can have cost-effective effects and impacts that go far beyond the initial design conceptualization.

The peer effect of health behaviors on middle-aged and older adult people is heterogeneous in terms of area of residence and level of education. Compared to urban communities, the health behaviors of middle-aged and older people in rural communities are more likely to be influenced by the average level of health behaviors in their community. Rural areas are natural communities, less subject to residential cluster effects and more conducive to the identification of health behavioral peer effects in middle-aged and older people (29). In addition, those who live in the countryside are more likely to form a society of acquaintances in which there are strong interactions and interdependence among middle-aged and older individuals, which reinforces the dissemination of health information (35). In terms of education level, the peer effect of health behaviors was the most effective in promoting health behaviors in middle-aged and older adult people with primary school education and below, slightly higher than that in the secondary school education group. However, there was no peer effect on the health behaviors of middle-aged and older adult people who had a tertiary education or above. There are three possible reasons for this. First, the middle-aged and the older adult with a lower level of education have fewer channels to obtain health information; the weaker their level of behavioral decision-making and judgment, the more likely that they will be influenced by role models in the process of social interaction, and the more sensitive they will be to the peer effect of health behaviors. Second, people with a high level of education usually have stronger cognitive and information comprehension and analytical and processing abilities, and are able to acquire health-related knowledge from multiple sources, resulting in a higher level of health literacy (36). As a result, their health behaviors are more autonomous in terms of decision-making and less affected by the peer effect of health behaviors. From the perspective of policymakers, rural communities and middle-aged and older groups with low levels of education should receive more policy-oriented support and assistance. For example, the role of health education should be brought into play to advocate for the adoption of positive health behaviors and lifestyles by the middle-aged and older adult; policy advocacy can be tailored toward specific groups of people, guiding cadres, party members, or opinion leaders with a high level of literacy to play the role of exemplars of health behaviors and give full play to the demonstrative effect of health behaviors.

Internet use played a significant and positive moderating role in the peer effect of health behaviors in the older population, i.e., Internet use increased the average health behaviors of older individuals. First, older persons have withdrawn from the labor market, with the consequent loss of the formal social roles and social relationships associated with work, so that their social networks are ultimately confined to their family and friends. Second, as physical function declines, so does the frequency of socializing and social activities among older people. Older adults who use the Internet extend their social network while staying in close contact with family and friends (37, 38). The stronger the social interaction, the more pronounced the peer effect of health behaviors. In addition, in the current context of widespread smartphone usage in China, the older adult can independently engage in searching for medical care and obtaining health information through the use of smartphones. Internet use provides a degree of independence for older people, freeing them from having to rely entirely on their children for personal care and attention (39). At the same time, Internet use helps older people obtain a wealth of health information and improve their disease prevention awareness; the resulting health literacy can help them develop good health behaviors (40). However, the spread of social media has also led to the proliferation of false health information. Intergenerational support from children can help older adults better use online health information, thus bridging the digital divide and promoting healthy behaviors (39, 41).

Meanwhile, our results also show that Internet proficiency played a significant negative moderating role in the peer effect of health behaviors on the middle-aged population, i.e., Internet proficiency reduced the impact of health behaviors on middle-aged individuals. There are two possible reasons for this. On the one hand, it has been argued that using devices to access information is a basic skill, and that the ability to recognize the value of information and quickly access needed information is particularly important (42). Unlike older people, middle-aged people have richer types of Internet online activities and more complex social networks, are able to distinguish and absorb information from different information channels, and are less likely to be influenced by a single group. On the other hand, the proficiency of middle-aged and older adults at Internet use implies that Internet addiction may be a problem (43). Substitution effects theory suggests that the time middle-aged people spend in online communities erodes their physical community interactions, which may diminish the impact of the peer effect. Therefore, in the process of continuing to improve the popularity of the Internet, the relevant departments (e.g., government agencies such as the Information and Communication Authority (ICMA), academic organizations such as Population Development and Gerontology Management (PDGM), age-friendly smart product development organizations, etc.) should provide guidance on reasonable use of the Internet for the middle-aged and older adult and give full play to the positive role of the Internet. For middle-aged adults living in the community, providing access to and training in digital technology can increase their self-efficacy with digital devices, thereby increasing the frequency of effective Internet-based communication with friends and family and reducing social isolation (44). At the same time, it is also necessary to make good use of the Internet as a carrier of health information and to enhance the older adult’s awareness of their health and their ability to take an active role in their own health through the transmission of health knowledge and concepts.

The peer effects of health behaviors on middle-aged and older adult people differed in terms of Internet use preferences. The peer effect of health behaviors on middle-aged and older adult people whose Internet preference was social and recreational was higher than that on those whose Internet preference was cultural and recreational, and the peer effect for those whose preference was cultural and recreational was higher than that for those whose preference was comprehensive, i.e., the peer effect of health behaviors was more sensitive among middle-aged and older adult people who used the Internet socially and instrumentally (i.e., for obtaining information, etc.).

First, middle-aged and older adult people whose preference is social entertainment go online mainly to watch videos and chat, and through chatting and short videos they enhance their interaction with friends and relatives and maintain social relationships (45). Short videos with social functions are simple to upload and rich in information, and thus can help middle-aged and older adult people share their lives and pay attention to the health conditions of their friends and relatives. Liking, commenting, forwarding, and interacting with each other on short video platforms can make up for the lack of offline communication, making the social distance between individuals closer and the peer effect of health behaviors stronger.

Second, middle-aged and older adult people with a cultural and recreational focus have active learning and self-entertainment online behaviors, mainly consisting of watching videos, reading news, playing games, and managing money. The Internet has the function of mass communication and interpersonal communication, making it possible for middle-aged and older adult people to learn a wealth of health knowledge through the Internet. The result is that the Internet changes the health behavior of the older adult through persuasive communication (46), and the peer effect is more likely to have an impact through the social learning and cohort mechanisms. It is worth noting that Internet videos have an advantage in Internet active learning among middle-aged and older adult people, in that, compared with static information, short videos are more vivid and intuitive, and are therefore more likely to be accepted by middle-aged and older adult people who have a low cultural level and limited reading ability (47).

Finally, the peer effect of health behaviors on middle-aged and older adult people whose Internet preference was of the comprehensive type was relatively low. The results of this study show that the proportion of middle-aged and older adult people with high Internet proficiency whose Internet preference was comprehensive was 46.99%, and their online activities included chatting, watching videos, reading news, playing games, and managing money. This indicates that middle-aged and older adult people with a comprehensive preference have rich online activities and high Internet networking proficiency and frequency of use.

There is some degree of Internet addiction and Internet fraud afflicting middle-aged and older adult people (48). The physiological functions of middle-aged and older adult people are in decline, and excessive use of the Internet will consume a certain amount of their energy and physical strength, but it is not conducive to the promotion of healthy behaviors, such that the peer effect is relatively weakened. This suggests that we should take advantage of the rapid development of the Internet and short videos as a form of communication in promoting health behavior change among the middle-aged and older adult. Older adults have lower cyber-efficacy and digital literacy, and thus they are less able to identify and defend against false health information. Under the Chinese culture of filial piety, intergenerational relationships can serve as a safety net for the older adult (41). High-quality intergenerational relationships can reduce the incidence of healthcare fraud among the older adult. There is a need to effectively address the problem of Internet fraud among the middle-aged and older adult by increasing publicity efforts to encourage children to pay attention to the emotional needs of their parents and increase the frequency of emotional communication with them (49).

Despite these important findings, this study has some limitations. First, it focuses on middle-aged and older adult individuals in the community under the Chinese socio-cultural context. Its results may not be generalizable to a wider population or a different economic and cultural environment. Second, the cross-sectional data limit our ability to draw any causal inferences about the impact of Internet use on the heterogeneity of peer effects on health behaviors. Third, the two dimensions of health behavior were based on a collation of indicators from the database. Therefore, the health behavior profile of respondents used in this study may not be comprehensive. Fourth, only three variables related to Internet use (namely Internet use, Internet proficiency, and Internet preference) were used due to the limitations of the available data. Further studies should consider adding other explanatory variables such as type, kind, and duration of Internet use. Finally, the data may be outdated because the latest round of CHARLS was updated in 2020.