Among older adults surveyed in Shandong, China, mobile social media functions did not show one clear relationship with either exercise identity or the amount of sedentary time tied specifically to social-media use. The overall paths in the statistical model were not supported for either outcome. But the picture changed when researchers separated social media into its uses: information retrieval, online guidance and social comparison were positively associated with exercise identity, while video media and personalized algorithms were negatively associated with it. The video and personalized-algorithms function was also positively associated with social-media-specific sedentary time. These are associations, not evidence that any function causes people to exercise more or sit longer.
One measure, several different patterns
Exercise identity means seeing exercise as part of who you are. The overall mobile-social-media-function construct had a point estimate of 0.72, but its corrected p-value was 0.36, so the combined direct path was not statistically supported. At the function level, the study reported positive paths for information retrieval, online guidance and social comparison, and a negative path for video media and personalized algorithms. The paths for attention and interaction, product service, goal setting and record keeping were not supported.
For sedentary time, the overall direct path was also unsupported, with a point estimate of 0.12 and a corrected p-value of 0.87. The exception was video media and personalized algorithms: its association with social-media-specific sedentary time was 0.50, with a 95% confidence interval from 0.17 to 1.01 and a corrected p-value of 0.04. The other function-specific paths were unsupported. The analysis was about sedentary time specific to social-media use.
The pathways were not the same
The researchers also tested two possible go-between pathways: perceived value and immersive, or flow, experience. Here, an indirect path means that a statistical association is linked through an intermediate measure. The overall path through perceived value was supported, with Z=3.34, a 95% confidence interval from 0.07 to 0.24, and a corrected p-value of 0.02. At the function level, supported indirect paths involved attention and interaction and record keeping. That pattern is consistent with perceived value being relevant to the exercise-identity association, but it does not show that making a function seem more valuable would cause a change.
Immersive experience showed a different result. Its overall indirect path was not supported, with Z=-0.92, a 95% confidence interval from -0.27 to 0.13, and a corrected p-value of 0.45. The video media and personalized-algorithms path was reported as supported, with Z=1.41, a 95% confidence interval from 0.006 to 0.19, and a corrected p-value of 0.05. Taken together, the results point to different statistical pathways for different social-media functions, rather than one uniform relationship.
How the evidence was assembled
The study used a cross-sectional survey in Shandong, with purposive recruitment followed by snowball sampling. Of 600 distributed questionnaires, 148 were excluded, leaving 452 participants and an effective response rate of 75.33%. The participants' mean age was 67.47 years, with a standard deviation of 6.04 years; 202 were male and 250 female, while 127 lived alone and 325 lived with others. The authors also reported ethics approval and written consent.
Researchers used structural equation modeling, a way to estimate several linked associations at once, alongside descriptive statistics and correlations. Correlation tests used false-discovery-rate correction, and the mediation tests used bias-corrected bootstrap calculations with 5,000 iterations. The feature measure covered eight constructs and 26 questions on a five-point scale; its McDonald's omega values ranged from 0.90 to 0.93. The reported model fit was judged acceptable, including a comparative fit index of 0.95 and a root mean square error of approximation of 0.04.
Why the findings remain tentative
The design sets clear limits on what can be concluded. Because the data were collected at one point in time, the analysis cannot establish causality or show which came first. Reverse causality remains possible: an older adult's existing exercise identity or social-media habits could be related to the functions they use. Purposive and snowball recruitment also limits how confidently the findings can be extended beyond the sampled older adults in Shandong and beyond China.
Several checks addressed the possibility that the questionnaire format was driving the results. A nested factor-analysis comparison produced a difference of 66 degrees of freedom and a chi-square difference of 9425.61, with p below 0.001; the authors concluded that common variance bias was absent. In a random-split sensitivity analysis, changes in measurement weights, structural weights, structural covariances and structural residual constraints were not statistically significant. Measurement residuals did differ, however, with a chi-square change of 157.63 and p below 0.001, although the authors still judged the model stable because its fit-difference indices met their standards.
Paper data and sources
Original title: A double-edged sword: unpacking the functional heterogeneity of mobile social media use and its differential associations with exercise identity and social-media-specific sedentary time among older adults
Authors: Liu N, Zhang G, Han X et al.
Journal/Repository: Frontiers in public health
Status: Peer-reviewed
First online: 2026-10-08
DOI: Not available
Original paper · Full text