A serial mediation model of physical exercise and loneliness: the role of frailty and depression | BMC Geriatrics

Recruitment and participants

This study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of Zhejiang Normal University (ZSRT20240125).

A purposive criteria-based sampling approach was employed to enhance the representativeness of the sample relative to the target population. Five cities in Zhejiang Province were selected for investigation based on population size, level of economic development, and regional spatial distribution. Specifically, Hangzhou was chosen as the most populous city in the province, Ningbo as the city with the highest per capita GDP, and Jinhua, Quzhou, and Wenzhou as cities representing distinct geographic regions within Zhejiang.

Before the surveys commenced, all participating researchers received training to ensure uniformity in the procedures and requirements across the surveys.

Prior to data collection, investigators coordinated with local sub-district offices and community neighborhood committees to recruit older adults who met the inclusion criteria. The inclusion criteria were as follows: (a) aged 60 years or older; (b) no language communication barriers or severe cognitive impairment, ensuring the ability to complete the questionnaire; and (c) provision of informed consent and voluntary participation. Exclusion criteria included: (a) serious mental or physical illnesses; (b) inability to read or understand the questionnaire; and (c) unwillingness to cooperate with the survey.

From February 25 to July 1, 2023, ten researchers were divided into two groups conducted a questionnaire survey among 600 older adults (aged ≥ 60 years) across the five selected cities. After excluding 95 questionnaires due to insincere responses, defined as those with identical answers for all questions, incomplete responses, or unclear answers, a total of 505 questionnaires were retained for final analysis.

Instruments

The questionnaire used in this study was developed in Chinese and consisted of three sections. First, participants were informed that the survey adhered to the principles of voluntariness and anonymity. They were assured that their responses would be accessible only to the research team and would not be used for commercial or non-research purposes. The second section collected demographic information about the older adults, including age, gender, education level, and other relevant background details. The third section included the standardized scales required for this study, as described below:

Physical activity rating scale (PARS-3)

The questionnaire was developed by Komio Hashimoto, translated in Chinese and revised by Liang Deqing et al. [41]. The questionnaire measured the level of physical activity in terms of intensity, time, and frequency. The PARS-3 Chinese version of the scale is evaluated using the 5-point scoring method and the scale consists of three questions. The total physical activity level was calculated by multiplying intensity by (time-1) by frequency, with 100 as the highest score and 0 as the lowest score. According to the theory of physical activity level, scores ≤ 19 are recorded as low exercise level, 20–42 are recorded as moderate exercise level, and ≥ 42 are recorded as high exercise level.

UCLA loneliness scale-8 (ULS-8)

The scale was adapted by Hays and DiMatteo based on the UCLA-20 scale (University of California Los Angeles Loneliness Scale), consisting of 8 items [42]. This study translated and revised it. The translated version has acceptable reliability (Cronbach’s α coefficient of 0.757) and validity (The fit indices: χ²/df = 4.577, CFI = 0.921, GFI = 0.959, RMSEA = 0.082, SRMR = 0.064) when used in the Chinese context. Among them, 6 items are positively stated as “loneliness” and 2 items are phrased negatively as “non-loneliness.” Each item is rated using a Likert 4-point scoring method, with scores of 1–4 indicating never, rarely, sometimes, and always. The non-loneliness items are reverse-scored. The score on this scale ranges from 8 to 32, where higher scores indicate higher levels of loneliness.

Tilburg frailty index (TFI)

TFI was developed by Gobbens et al. [16], translated in Chinese and revised by Si et al. [43]. The scale was developed on the basis of the integral conceptual model of frailty, and measured the frailty status of the older adults from three dimensions: physical, psychological and social frailty, with a total of 15 items. The scale items adopt the two-classification scoring method, and the scoring range is 0 ~ 15 points, and 5 points and above points are frailty. The higher the score, the heavier frailty. The scale demonstrates good reliability with a Cronbach’s α coefficient of 0.787.

Patient health questionnaire (PHQ-9)

PHQ-9 based on DSM-IV is used to evaluate depression symptoms, comprising nine items. This questionnaire was translated and revised by Chen and has been widely used in the Chinese context [44]. The scoring criteria for PHQ-9 scale were proposed by Kroenke [45]. Each item is scored from 0 to 3, with 0 indicating not at all, 1 indicating several days, 2 indicating more than half the days, and 3 indicating nearly every day. The total score ranges from 0 to 27, with cutoff points at 5, 10, 15, and 20, categorizing the severity of depression as follows: no depression (0–4), mild (5–9), moderate (10–14), moderately severe (15–19), and severe (20–27). Clinically, a score of ≥ 5 indicates a potential association with significant depressive symptoms, with higher scores indicate more severe depression. The scale shows good reliability with a Cronbach’s α coefficient of 0.865.

Data collection

Informed consent was obtained from all participants prior to their involvement in the study. The purpose of the research, along with assurances of confidentiality and anonymity, was clearly explained to ensure participants were fully aware of the procedures and any potential risks. Participants were informed of their right to withdraw from the study at any time without consequence.

The study was conducted as an anonymous survey, and all data collected were used solely for scientific research purposes. To minimize external interference and ensure accurate responses, researchers, accompanied by community staff, visited the homes of older adults to administer the questionnaires. Researchers guided participants throughout the questionnaire completion process, assisting them in understanding the meaning of the items. For participants who had difficulty reading or comprehending the questions, researchers read the items aloud and provided additional explanations as needed to facilitate accurate responses.

On average, participants took approximately 20 min to complete the questionnaire. All questionnaires were collected immediately after completion to ensure data integrity.

Data analysis

Data processing and statistical analysis were conducted using SPSS 26.0 (IBM, Armonk, NY, USA) and AMOS 27.0 (IBM, Armonk, NY, USA). Independent sample t-tests and one-way ANOVA were performed to compare group differences. To address potential common method bias, Harman’s single-factor test was applied. Pearson correlation analysis was used to examine relationships between variables. Regression analysis and mediation effects were tested using the bias-corrected percentile bootstrap method.

Mediation effects were tested using AMOS. AMOS is a software designed for SEM. It enables the simultaneous estimation of relationships among multiple variables, including direct, indirect, and total effects. For confirmatory factor analysis, AMOS can intuitively construct factor models and provide rich fitting indices to help us evaluate the fitting degree of the model, which is crucial for our research. The Bootstrap method was employed. The parameter settings as follows: the number of bootstrap samples was set to 5,000 to ensure the stability and accuracy of the estimates, the confidence level was set to 95% for calculating the confidence intervals of the mediation effects. In the output, the primary statistical values of interest included the estimated mediation effect, standard error, t-value, and 95% confidence interval. These values are reported in the Results.

A significance level of p < 0.05 was set for all statistical tests.

Reliability and validity

Confirmatory Factor Analysis (CFA) was employed to assess the internal validity of each construct. For the ULS-8, the fit indices were as follows: χ²/df = 4.577, CFI = 0.921, GFI = 0.959, RMSEA = 0.082, SRMR = 0.064. While the chi-square value can be influenced by sample size, value less than 3 are considered ideal, and values less than 5 are relatively acceptable. For the TFI, the fit indices were χ²/df = 1.340, CFI = 0.970, GFI = 0.968, RMSEA = 0.026, SRMR = 0.038, and the PHQ-9 were χ²/df = 1.103, CFI = 0.999, GFI = 0.989, RMSEA = 0.014, SRMR = 0.018. All meet the validity score standards for scales (χ²/df < 3, CFI > 0.9, GFI > 0.9, RMSEA < 0.08, SRMR < 0.08), indicating a good fit for these scales. In terms of reliability, the Cronbach’s α coefficients for the ULS-8, the TFI, and the PHQ-9 were 0.757, 0.787, and 0.865 respectively. Both Cronbach’s α coefficients and composite reliability (CR) exceeded the threshold of 0.7, confirming good reliability for the mentioned scales.

Common method bias test

Since all structural data were collected through self-reported surveys, common method bias (CMB) is a potential concern. To mitigate the impact of CMB on the validity of the results, several procedural measures were implemented during the survey process, including balanced item ordering, anonymous questionnaire administration, and standardized measurement protocols. Additionally, Harman’s single-factor test was conducted to assess the extent of CMB. The results indicated that the first principal component accounted for 24.47% of the total variance, which is below the recommended threshold of 40%, suggesting that common method bias was not a significant issue in this study.