This secondary analysis used baseline data collected from an ongoing cohort study named the Physical Activity and Health in Older Women Study (PAHIOWS) launched from March to June 2021 in Yantai, Shandong Province, China. All investigators in PAHIOWS were trained to collect self-reported and objective data in advance by the principal investigator. Following the gatekeeper’s (ie, the director of the community center) approval, the recruitment of participants was launched. Traditional approaches including poster campaigns and person-to-person interactions were used to recruit participants. To entice participation, both monetary (gifts) and nonmonetary incentives were used. The nonmonetary incentive was a printed health report including the information on body composition assessed by Tanita MC-180 (Bailida Co, Tokyo, Japan) and arterial stiffness and vascular obstruction assessed by an automatic oscillometer device (VaSera VS-1500AE, Fukuda-Denshi, Tokyo, Japan). The inclusion criteria set for the parent study were community women ages 60-70 years who could communicate, had no cognitive impairment (the Mini-Mental State Examination score: >17 for illiteracy, >20 for primary school, and >24 for middle school and above), and provided written informed consent. There were 1370 older women enrolled in PAHIOWS; all data were collected in person, and a double-checking strategy was used to eliminate data entry mistakes. In our study, we solely used data from participants using smartphones with short-form video apps Tiktok/Douyin or Kuaishou, Toutiao, and Haokan, as they were apps older adults often used in China.

The ethical oversight of the parent study was obtained from the institutional review board in the School of Nursing and Rehabilitation, Shandong University, China (2020-R001).

No outlier was detected by visualizing the data with a box plot, and no strategy was used to handle the missing values in this study as the number of missing values was extremely small (around 1%). Sample characteristics were presented as means (SDs) or medians (IQRs) for continuous variables and presented as numbers (percentages) for categorical variables. The content exposure and the duration of exposure were regressed on step counts, LPA, and MVPA, respectively. Covariates including age, education background, living alone, BMI, history of falls, history of heart diseases or otolithiasis, balance index level, gait speed, musculoskeletal pain, and forced vital capacity were controlled for each regression model. BMI, education background, and the balance index level were transformed into dummy variables before entering each linear regression model. For MVPA, we did an additional analysis by dichotomizing MVPA into below 150 minutes per week and 150 minutes and above per week, and regressed content exposure and the duration of exposure on it while controlling for all covariates. We also did sensitivity analyses by replacing the history of falls with the history of falls with injuries in all models. The unstandardized regression coefficient (B), odds ratio (OR) and its 95% CI were calculated. All analyses were performed with SPSS (version 26; IBM Corp), and P<.05 was considered statistically significant.

Table 1 summarizes the characteristics of the sample. Older women in this secondary analysis had a mean age of 64.63 (SD 2.90) years, with the majority receiving education at the middle and high school level (n=351, 73.7%), living with others (n=426, 89.5%), and having abnormal BMI (n=377, 79.2%). Around one-third of participants had a history of falling (n=133, 27.9%) or a history of heart disease or otolithiasis (n=157, 33%), and 24.6% (n=117) of participants reported a history of falling with injuries. Most of the 476 participants were graded as low on the balance index level (n=383, 80.5%) and had forced vital capacity lower than 2500 ml (n=392, 82.4%). Approximately two-thirds of participants had musculoskeletal pain (n=310, 65.1%), and as many as 97.1% (n=462) of participants in this study demonstrated satisfactory gait speed (ie, ≥1 m/s). There were 23.7% (n=113) of participants exposed to PA-related short-form videos, and the median duration of exposure to short-form videos was 1.5 (IQR 1-2.5) hours per day. Participants walked an average of 8186.84 steps per day monitored by ActiGraph wGT3X-BT, and the mean (SD) time spent on LPA and MVPA was 303.15 (76.60) and 32.11 (18.83) minutes per day, respectively. Only a few participants (n=71, 14.9%) in this study spent 150 minutes or more on MVPA weekly.

Table 2 presents the results of linear regression models on step counts, LPA, and MVPA. When holding other variables as constant, a 1-unit increase in the duration of exposure to short-form videos significantly decreased 322.58 walk steps per day for older women. We also found significantly negative impacts of the duration of exposure to short-form videos on the minutes spent on the LPA (B=−6.95, 95% CI −12.19 to −1.71) and MVPA (B=−1.56, 95% CI −2.82 to −0.29) of older women. PA-related content exposure was only found to be significantly associated with increases in the minutes spent on MVPA (B=4.14, 95% CI 0.13-8.15). In the binary logistic regression model, the content exposure failed to significantly increase the odds of spending 150 minutes or more on MVPA (OR 1.313, 95% CI 0.723-2.381), and the duration of exposure failed to significantly reduce the odds (OR 0.949, 95% CI 0.772-1.167). The results remained stable when replacing the variable of history of fall with history of fall with injuries in each model.

As depicted in Table 2, we found LPA reduced with advancing age (B=−3.70, 95% CI −6.20 to −1.19). A history of falling significantly reduced daily steps (B=564.57, 95% CI −1102.76 to −26.39) for older women, but failed to significantly reduce the minutes spent on LPA and MVPA. When replacing the history of fall with the history of fall with injuries, we found the history of fall with injuries significantly reduced daily steps (B=−586.44, 95% CI −1146.69 to −26.20). A BMI of 28 and above was significantly associated with increases in the minutes older women spent on LPA (B=27.58, 95% CI 7.55-47.61), and forced vital capacity of 2500 ml and above was found to be significantly associated with increases in the minutes spent on MVPA (B=5.29, 95% CI 0.88-9.70). In the binary logistic regression model, forced vital capacity of 2500 ml and above significantly increased the odds of spending 150 minutes or more on MVPA (OR 2.247, 95% CI 1.224-4.125), while a history of falling significantly reduced the odds (OR 0.423, 95% CI 0.204-0.877). Significant results remained stable when replacing the history of falls with the history of falls with injuries in each model.

Quarantine, limited in-person gatherings, and other lockdown measures as a result of COVID-19 have restricted PA [32,33] and simultaneously increased the number of older adult netizens [15]. Approximately one-third of new older adult netizens were short-form video users, and short-form video apps have demonstrated higher user stickiness, indicating that users would stay longer and frequently revisit [16]. Furthermore, empirical evidence from younger adults demonstrated that social media exposure influences their health behaviors such as food choice and smoking behavior [25,34]. However, less is known about the status of short-form video exposure and its association with PA among older adults. By analyzing 476 older women who were users of short-form video apps, we found that 23.7% (n=113) of them were exposed to PA-related short-form videos, and their daily exposure to short-form videos was 1.5 hours. In addition, we identified that the duration of exposure was significantly associated with reduced step counts and minutes engaged in LPA and MVPA. PA-related content exposure was significantly associated with an increase in minutes spent on MVPA but failed to significantly increase the odds of engaging in MVPA for 150 minutes or more per week among older adults.

Evidence has demonstrated that taking at least 7500 steps daily generates health benefits. For example, Saint-Maurice et al [35] analyzed data collected from 4840 adults (mean age 56.8 years; n=2435, 54% women) and found that taking 8000 steps per day was associated with significantly lower all-cause mortality compared with taking 4000 steps per day. Lee et al [7] analyzed data collected from 16,741 older women (mean age 72 years) and found that mortality rates progressively decreased before leveling at approximately 7500 steps per day. In this study, older women took an average of 8186.84 steps daily (ie, exceeding 7500 steps), and we found that a 1-hour increase in short-form video exposure significantly reduced 322.8 steps per day. This finding indicates that research is needed to explore the appropriate time spent on short-form videos by older women, and such efforts might inform strategies for app developers and community health workers to avoid the hazardous impacts of short-form videos on the health of older women.

We found that women spent an average of 303.15 minutes per day on LPA and 32.11 minutes per day on MVPA, and approximately 15% of older women would achieve the goal of spending 150 minutes or more on MVPA per week, which was slightly higher than the rates reported in studies from high-income countries [36,37]. The longer duration of exposure was significantly associated with a reduction in the minutes spent on LPA and MVPA, and the PA-related content exposure was significantly associated with an increase in the minutes spent on MVPA. Neither the duration of exposure nor PA-related content exposure was found to be significantly associated with older adults’ engagement in MVPA for 150 minutes or more per week. Being physically active is beneficial for healthy aging [1,2]; our findings strengthen the necessity of investigating the recommended time spent on short-form video apps. In addition, we found that less than one-quarter of older women viewed short-form videos related to PA. Short-form video apps often recommend content starting from the interests or preferences of the users; our findings indicate that short-form video apps may optimize the recommendation system such as passively pushing PA-related short-form videos, and such efforts may increase the content-related exposure of older women and promote engaging in more MVPA.

Except for media exposure, we found several sample characteristics that were associated with PA in older women. In this study, 27.9% (133/476) of older women reported a history of falls, which was slightly lower than those reported in older adults of both genders (ie, 30%-50%) [38], and women with a history of falls would take 564.57 fewer steps per day than those without, reducing the odds of achieving the goal of spending 150 minutes or more per week on MVPA. We also found that older women with a BMI of 28 kg/m² would spend an extra 27.58 minutes per day on LPA compared with those with normal BMI, and this is inconsistent with the evidence in the literature that PA declines with an increase in weight [39]. This inconsistency can be explained by the historical effect of COVID-19 (ie, obesity is a well-known risk factor for COVID-19, severe COVID-19, and its complications, and when sparked by the pandemic, older women with a BMI above 28 kg/m² may have become more sensitive to their health and formed new behaviors such as being physically active to protect or enhance their well-being [40,41]). Last, forced vital capacity declined by 20-30 ml per year starting at the age of 30 years [42], and in this study, only 17.6% (84/476) of older women had forced vital capacity of 2500 ml or more. Exercise limitation is a well-known consequence of respiratory conditions [43], our study corroborated this notion and found that compared with women with unsatisfactory forced vital capacity, women with a forced vital capacity ≥2500 ml were more likely to have spent more minutes on MVPA and to achieve the goal of engaging in MVPA for 150 minutes or more per week.

This study is not free of limitations. First, limited by the secondary data analysis, we failed to assess and control variables that might influence the inference of our study. For example, older adults are accustomed to using traditional media [44], which may also transfer PA-related information and influence the outcome of interest in this study. Additionally, the built environment is particularly relevant to the PA of older adults [45,46], and this variable was not included in PAHIOWS either. Future studies may want to corroborate findings from this study by assessing and controlling these variables. Second, we failed to document the time older women spent specifically watching PA-related short-form videos. The duration of time older women spent watching PA-related short-form videos may provide a better understanding of the impact of content exposure on PA. However, collecting and calculating PA-related short-form video exposure would be challenging; future studies may want to develop strategies to collect these data and that may help give more specific recommendations for the PA of older adult netizens. Third, we found opposite conclusions on relationships between content exposure with PA and the duration of exposure with PA. Moreover, empirical evidence has demonstrated that media exposure would bring mental health benefits to older adults [47]. More studies are needed to explore the balance point to leverage the benefit while avoiding the negative effects of short-form videos. Lastly, PAHIOWS recruited only community older women aged 60-70 years; extrapolating the findings from this study to a sample of male and female older adults or different age cohorts should be done cautiously.

Our study found that short-form videos, which are being increasingly viewed by older adults, introduced both positive and negative impacts on their PA. Exposure to content related to PA would increase the minutes older adults spent on MVPA per day, while the duration of exposure would decrease their steps counts and minutes spent on LPA and MVPA. However, neither of the exposures were able to predict the achievement of spending 150 minutes or more per week on MVPA among older women.