This study employed a prospective longitudinal design and used purposive sampling to recruit older adults participants who met the following inclusion criteria from communities, nursing homes, and senior universities in Guiyang and Duyun cities in Guizhou Province, as well as Nanchang city in Jiangxi Province: (1) Age between 60 and 80 years; (2) No severe cognitive impairments, able to understand and complete the questionnaires; (3) Residing in the specific cities or communities defined for this study to ensure regional representativeness. In each setting, local community workers, nursing home administrators, and senior university staff assisted in identifying potentially eligible older adults and informing them about the study. Recruitment information was disseminated through posters, notice-board announcements, and brief on-site presentations, and interested individuals were approached face to face by trained research assistants. The research assistants provided a concise introduction to the study aims and procedures, conducted an initial eligibility screening based on the inclusion criteria, and then invited those who were eligible to participate. Older adults who agreed to take part signed a written informed consent form before completing the questionnaire.

The baseline survey was conducted from February to April 2024, with a total of 1,068 valid questionnaires collected. The follow-up survey was completed between February and April 2025, with 967 older adults participants who participated in both rounds of the survey, meeting the above inclusion criteria, and were included in the final analysis. To address missing data, this study employed multiple imputation based on chained equations in the R programming environment, using the mice package to impute missing values (40). To further examine potential biases due to sample attrition, we compared participants who completed both rounds of the survey with those who only completed the baseline survey. Independent samples t-tests showed no significant differences between the two groups in terms of gender (t = 0.417, p > 0.05), marital status (t = −0.878, p > 0.05), education level (t = −1.202, p > 0.05), annual income (t = 0.401, p > 0.05), baseline physical activity levels (t = −1.960, p > 0.05), or psychological well-being (t = 1.920, p > 0.05). These results indicate that sample attrition did not lead to structural bias.

The final sample included 967 older adults individuals who completed both rounds of assessments, with their basic characteristics shown in Table 1. Among them, 372 were male (38.5%) and 595 were female (61.5%). The age distribution was primarily between 60 and 70 years, with 536 individuals (55.4%), followed by 321 individuals (33.2%) aged 71 to 80, and 110 individuals (11.4%) aged over 80. In terms of education level, 108 individuals (11.2%) were illiterate or semi-illiterate, 297 individuals (30.7%) had completed primary school, 398 individuals (41.2%) had completed middle school, and 164 individuals (17.0%) had completed high school or higher. Regarding marital status, 746 individuals (77.1%) had a spouse, while 221 individuals (22.9%) were without a spouse. As for annual income, 593 individuals (61.3%) had an income below 7,500 RMB, while 374 individuals (38.7%) had an income above 7,500 RMB.

This study employed Harman’s single-factor test to analyze common method bias. The results showed that in both time points, there were seven factors with eigenvalues greater than 1 (47). The first factor explained 35.45% of the variance at T1 and 37.52% at T2, both of which were below the 40% critical threshold. This indicates that there is no serious common method bias issue in the data of this study.

To comprehensively describe the sample characteristics, Table 2 presents the descriptive statistics for the indicators of physical activity and psychological well-being at both time points. Based on this, one-way analysis of variance (ANOVA) and independent samples t-tests were conducted to examine the influence of demographic variables on the core study variables. The results indicated that gender, age, education level, marital status, and annual income all had significant effects on physical activity levels and psychological well-being scores. Therefore, in the subsequent network analysis, these variables were included as covariates in the model for control.

The results of the cross-sectional network analysis showed that the network at the T1 time point consisted of 108 non-zero edges, with a network density of 0.568, indicating that 56.84% of possible connections were actually present. By the T2 time point, the complexity of the network increased, with the number of non-zero edges rising to 114, and the network density increasing to 0.600 (60.00%), reflecting an overall enhancement in connection strength. A comparison of the network structures between the two time points (Figures 1, 2) revealed a trend of strengthening internal associations over time. In the T1 network, the strongest association appeared within psychological well-being, specifically between “I believe life is a continuous process of learning, change, and growth” (PWB7) and “I am open to new things and new experiences” (PWB8) (r = 0.488). The most prominent cross-domain connection was between “How many minutes do you engage in the above-intensity physical activity at a time?” (PA2) and “When I need it, I can receive emotional support and help from my friends” (PWB12) (r = 0.048), suggesting a weak but identifiable link between physical activity and social support.

By T2, the strongest association in the network shifted to “I live according to my own beliefs and values, regardless of what others say” (PWB2) and “I am confident in my viewpoints, even if they differ from the majority” (PWB3) (r = 0.393), reflecting the stable covariance between individual autonomy and self-assurance. The key cross-domain edge at this point was between “How intense is your exercise?” (PA1) and “Over time, I can see myself making progress and maturing” (PWB9) (r = 0.061), indicating the emergence of a preliminary predictive relationship between exercise intensity and a sense of personal growth. The specific values of all the edges in the cross-sectional networks are provided in Supplementary Tables S3, S4. Overall, the network structure at both time points maintained stable connections between core psychological constructs and also demonstrated the gradual emergence of dynamic relationships between physical activity and the psychological well-being sub-systems.

Figure 3 shows the expected influence (EI) centrality values of the two cross-sectional networks at T1 and T2. At the T1 time point, the node with the highest centrality was “In retrospect, I am satisfied with my life experiences” (PWB18, EI = 1.740), indicating that life satisfaction had a broad influence within the system. By the T2 time point, the most influential node in the network shifted to “I believe life is a continuous process of learning, change, and growth” (PWB7, EI = 1.571), reflecting the core role of a growth mindset in the later network. Both of the most central nodes at T1 and T2 belonged to the psychological well-being dimension, reflecting its dominant position in the system structure. Specific centrality indicators for all nodes at both time points are detailed in Supplementary Tables S5, S6.

To assess the reliability of the cross-sectional network structure, this study used the Bootstrap sampling method to test the stability and accuracy of the T1 and T2 networks. The results showed that the 95% Bootstrap confidence intervals for the edge weights in both networks were relatively concentrated, indicating that the network estimates were overall stable. Regarding the correlation stability coefficients, in the T1 network, the CS coefficient for edges was 0.75, showing high stability; the eigenvector centrality was 0.517, reflecting good stability; the expected influence and strength centralities both had a CS of 0.361, which is within an acceptable range. In the T2 network, the stability of the indicators further improved: the CS for edges reached 0.75, representing the best stability; the eigenvector centrality was 0.672, demonstrating good stability; and the expected influence and strength centralities were 0.438, ranging from acceptable to good levels. All CS coefficients were above the minimum acceptable threshold of 0.25, indicating that the network centrality estimates had sufficient stability to support further analysis. The Bootstrap confidence intervals for the edge weights and the results of centrality difference tests are detailed in the supplementary materials.

The cross-lagged panel network included 166 non-zero edges, revealing the complex temporal interactions between physical activity and psychological well-being across various dimensions (Figure 4). Among all the predictive paths, the strongest cross-lagged effect was observed in the physical activity dimension, where “exercise duration” (PA2) had the most significant predictive effect on “exercise frequency” (PA3) (β = 0.187), indicating that the time invested in exercise can effectively predict the subsequent maintenance of exercise frequency. Within psychological well-being, “self-acceptance” (PWB17) exhibited a significant predictive effect on “self-affirmation” (PWB16) (β = 0.184), reflecting the role of self-acceptance in promoting overall self-evaluation. Importantly, the study also found a cross-domain predictive path from physical activity to psychological well-being, specifically with “exercise frequency” (PA3) having a significant positive predictive effect on “autonomous values” (PWB2) (β = 0.166), suggesting that regular exercise participation may enhance individuals’ sense of value autonomy. In terms of autoregressive effects, “exercise duration” (PA2) showed the strongest self-predictive effect (β = 0.152), demonstrating the relative stability of this behavior over time. The specific coefficients for these core predictive paths can be found in Supplementary Table S7.

The results of the centrality analysis of the cross-lagged panel network are shown in Figure 5. In terms of predictive influence, the nodes with the highest outward expected influence (OEI) were “exercise frequency” (PA3, OEI = 1.088) and “meaning in life” (PWB14, OEI = 0.830), indicating that these two indicators have a broad predictive effect on other variables within the system. Specifically, PA3, as the core behavioral indicator of physical activity, demonstrates a significant driving potential for the psychological well-being system. Meanwhile, PWB14 reflects the central role of meaning-seeking in promoting the positive development of the system. In terms of receiving influence, “self-affirmation” (PWB16, IEI = 0.723) and “meaning in life” (PWB14, IEI = 0.662) exhibited the highest inward expected influence (IEI), suggesting that they are more likely to be influenced by other variables and are important targets of influence within the network. Notably, PWB14 ranked highly on both OEI and IEI, indicating that it holds both a driving and sensitive role, and it may play a key bridging role in the dynamic interaction between physical activity and psychological well-being. The centrality indicators for all nodes can be found in Supplementary Table S8.

The stability test of the cross-lagged panel network showed that the correlation stability coefficient (CS coefficient) for the network edge weights was 0.283, which meets the acceptable standard (CS > 0.25), confirming that the network structure estimates have basic reliability. The detailed Bootstrap confidence intervals are provided in the supplementary materials.

The age distribution of participants in this study indicated that more than 50% of the participants were between the ages of 60 and 70 years. This age group may correspond to the typical retirement age in China, where the legal retirement age is 60 for men and 55 for women. At the T1 time point, the strongest association within the psychological well-being dimensions was observed between “personal growth” (PWB7) and “openness to new experiences” (PWB8), which aligns closely with psychological well-being theory. This suggests that, in old age, continuous learning and cognitive flexibility remain core elements of psychological flourishing (48). In the cross-domain connections, the weak link between “exercise duration per session” (PA2) and “social support availability” (PWB12) may be limited in strength but hints that physical activity could provide emotional support resources for older adults individuals through social interaction mechanisms (49). This aligns with the social support theory which posits that participation in activities can promote social connections (50). By T2, the core association in the network shifted to the high covariance between “autonomous value” (PWB2) and “self-assurance” (PWB3), indicating that, over time, the unity and stability of an individual’s internal value system became the center of the psychological well-being system (51). This result supports the core position of autonomy as a fundamental psychological need in self-determination theory (18), and also reflects the increasing importance that older adults individuals in China place on internal value recognition as they adapt to aging (52). At the same time, a preliminary connection emerged in the cross-domain associations between “exercise intensity” (PA1) and “sense of personal growth” (PWB9), suggesting that the intensity of physical activity may begin to be perceived by individuals as a positive experience that promotes self-development (53). This also indicates that high-intensity exercise may have potential value in fostering psychological growth (54).

Of particular note is the significant shift in the cross-domain associations from T1 to T2: from the initial weak connection between physical activity and social support, to a substantial association between exercise intensity and a sense of personal growth. This evolution reveals a potential stage-specific characteristic in the internal mechanisms by which physical activity influences psychological well-being—during the initial phase, it primarily works through enhancing social support networks, and as participation deepens, it increasingly promotes psychological well-being by stimulating personal growth experiences (55). This finding has important implications for practical interventions: different strategies should be adopted at various stages in promoting the physical and mental health of older adults individuals. In the early stages, the focus should be on establishing social support through group activities, while in the later stages, attention should shift to how personalized exercise prescriptions can stimulate individual growth experiences, thereby achieving more precise and effective health promotion (56, 57).

The centrality analysis of the cross-sectional network revealed the temporal evolution of key driving factors in the physical and mental system of older adults individuals. At the T1 time point, “life satisfaction” (PWB18) exhibited the highest centrality, which aligns with the stage characteristics of psychological adaptation in old age (58). According to psychosocial development theory, the primary developmental task in old age is achieving self-integration, and life satisfaction is a core indicator of an individual’s accomplishment of this task (59). Early in the aging process, individuals are more likely to establish the meaning of life by reviewing and evaluating their past life experiences, making life satisfaction a key node influencing the entire psychological well-being system (60).

By the T2 time point, the most influential node in the network shifted to “growth mindset” (PWB7), reflecting a shift in the focus of psychological adaptation in older adults individuals (61). As individuals gradually complete the integration of the past, their attention may shift from “retrospective evaluation” to “prospective development.” According to the multidimensional model of psychological well-being, personal growth is a fundamental element throughout life, and maintaining a growth mindset in old age is particularly important (62). It helps older adults individuals maintain a positive mindset and adaptability when facing physiological decline and changes in social roles (63). This shift also supports the theory of successful aging in the SOC (Selection, Optimization, and Compensation) model, which suggests that older adults individuals adjust the allocation of psychological resources, shifting the focus from simple satisfaction to a more vibrant sense of growth (64).

The temporal changes in centrality indicators reveal the dynamic characteristics of the psychological adaptation process in older adults individuals: from the initial phase of integration and acceptance to the later phase of development and growth (65). This finding has important implications for practical interventions, suggesting that when designing health promotion programs for the older adults, differentiated strategies should be adopted based on the psychological needs at different stages. In the early stages of the program, the focus should be on helping older adults individuals establish a positive life review and acceptance (66). In subsequent stages, the emphasis should shift to cultivating a mindset of continuous learning and growth, thereby achieving more precise and effective psychological enhancement (67).

The cross-lagged panel network analysis revealed the complex temporal interaction mechanisms between physical activity and psychological well-being. First, within the physical activity system, the predictive effect of “exercise duration” (PA2) on “exercise frequency” (PA3) was the most significant, which aligns with the theory of behavioral consistency (68). Empirical research shows that once older adults individuals invest longer durations in a single exercise session, they are more likely to develop regular exercise habits (69). People tend to maintain established behavior patterns to reduce decision-making costs. In the context of community activities in China, longer exercise durations often signify deeper social interactions and skill mastery, which together help sustain exercise habits (70). Second, within the psychological well-being system, the predictive path from “self-acceptance” (PWB17) to “self-affirmation” (PWB16) profoundly reflects the internal mechanisms between dimensions in the psychological well-being model (71). Empirical studies show that older adults individuals’ acceptance of their limitations is fundamental to building a stable sense of self-worth (72). Especially in the context of traditional Chinese culture, which emphasizes the concept of “knowing one’s destiny” (Zhi Tian Ming), older adults individuals’ full acceptance of themselves is more likely to translate into positive self-evaluation (73). Third, the cross-domain prediction from “exercise frequency” (PA3) to “autonomous values” (PWB2) empirically confirms the prospective association of physical activity on psychological development. This finding is consistent with the core proposition of self-determination theory, which suggests that the satisfaction of basic psychological needs has domain-general characteristics (18). Observations in the field show that older adults individuals who frequently participate in group activities such as Tai Chi or square dancing not only improve their physical fitness through skill mastery and rule adherence, but also enhance their ability to make independent decisions and value judgments (74, 75). Finally, the significant autoregressive effect of “exercise duration” (PA2). Research indicates that exercise duration, as a specific and measurable behavior indicator, is more likely to form stable behavior patterns compared to other physical activity dimensions (76). This characteristic is particularly evident in the context of Chinese health culture, which emphasizes “persistence” in maintaining a consistent and sustained approach to health practices (77).

These findings collectively outline a virtuous cycle for promoting the physical and mental health of older adults individuals: by establishing stable exercise patterns, psychological basic needs are met, which in turn promotes the overall development of psychological well-being. This mechanism provides an important theoretical foundation and practical guidance for designing phased, multi-level health intervention programs for the older adults.

The cross-lagged network centrality analysis further revealed the dynamic roles and functional positioning of various elements within the physical activity and psychological well-being systems. The results showed that “exercise frequency” (PA3) exhibited the strongest outward predictive influence (OEI = 1.088), which dynamically supports the core concept of habit formation theory (78). Empirical research indicates that regular participation in physical activity is not only an indicator of healthy behavior but also a key driving force that initiates the positive cycle of mind and body (79). In the life context of older adults individuals in China, fixed weekly group activities such as square dancing and Tai Chi help establish a regular life rhythm and social participation pattern, providing stable external input to the entire psychological well-being system (80).

A particularly important aspect to explore is the special position of “meaning in life” (PWB14) within the system, as it holds both a high outward predictive influence (OEI = 0.830) and a significant inward influence (IEI = 0.662). This dual characteristic aligns with the core concept of Frankl’s meaning-centered therapy theory, which emphasizes that the search for meaning is both a goal of psychological development and a process that promotes growth (81). In the Chinese cultural context, older adults individuals often seek meaning in life through various avenues, such as intergenerational transmission and community participation, with regular physical activity providing a platform and environment for realizing these meanings (82). The bridging role of PWB14 suggests that it may be an important hub connecting physical activity and the dimensions of psychological well-being—both transforming the benefits of physical activity into deeper psychological experiences and further promoting the depth and continuity of exercise participation through an enhanced sense of meaning (83). On the other hand, “self-affirmation” (PWB16), as the most inwardly influenced node (IEI = 0.723), reflects the acceptance characteristics of older adults individuals in the mind–body interaction process. In the identity transition that Chinese older adults individuals experience after retirement, the establishment of self-affirmation often depends on continuous external recognition and achievement experiences, and physical activity provides a platform for such validation (84).

The distribution characteristics of these centrality indicators provide important insights for designing targeted health promotion programs for the older adults: enhancing meaning in life should be prioritized as the core intervention target. By creating physical activity programs that are rich in meaning, it is possible to simultaneously activate older adults individuals’ behavioral engagement and psychological development, thereby establishing a virtuous cycle where physical activity and psychological well-being mutually promote each other (85).

Finally, when considering the future social context of aging, it is important to recognize that older adults are increasingly likely to live in a hybrid society in which human relationships are intertwined with interactions with AI agents, including systems based on large language models. Recent studies indicate that such agents can display culturally patterned values, interpersonal styles, and biases, and that their behavior may vary across cultural settings (86). If AI companions, digital health coaches, or service robots become integrated into older adults’ daily lives, they may alter the pathways identified in this study—for example, by shaping access to physical activity opportunities, providing new forms of social and emotional support, or reinforcing particular value orientations related to autonomy and meaning in life (87). Future longitudinal research should therefore examine how AI-mediated interactions interface with the network of physical activity and psychological well-being, and whether differences in the design and cultural alignment of AI systems strengthen, weaken, or qualitatively change the dynamic associations observed here for Chinese older adults.

This study, by introducing the cross-lagged panel network analysis method, overcomes the limitations of traditional variable-centered paradigms and, for the first time, reveals the multidimensional and bidirectional interaction paths between physical activity and psychological well-being in older adults individuals within a dynamic system. From a theoretical perspective, this research not only validates the core role of basic psychological needs in the mind–body connection, as proposed in self-determination theory (18), but also expands the explanatory scope of the successful aging model in understanding the behavioral-psychological interaction mechanism (88). The study highlights the bridging role of meaning in life within the system, providing a new theoretical perspective for understanding the psychological adaptation process of older adults Chinese individuals from social connection to value realization. Furthermore, it provides empirical evidence for the development of psychological health frameworks for the older adults in cross-cultural contexts.

On a practical level, this study provides scientific evidence for designing targeted, phased health promotion programs for the older adults. The results suggest that in the early stages, social support should be strengthened through group physical activities, while later stages should focus on stimulating personal growth experiences (89). This finding aids communities and service organizations in developing more targeted mind–body integration intervention strategies. Moreover, with exercise frequency and “meaning in life identified as key driving nodes, the study emphasizes the importance of incorporating meaning-rich physical activity programs in public health projects, effectively enhancing participation continuity and psychological benefits. These outcomes have significant practical implications for promoting healthy aging, moving from concept to implementation, and optimizing the older adults health service system.