A theory-driven exercise intervention to improve physical performance in community-dwelling older adults: a cluster randomized controlled trial delivered by general practitioners

Abstract

Background

Age-related physical decline demands effective community-based interventions. Theory-driven programs, especially those led by general practitioners (GPs) and integrating behavior change techniques (BCTs), remain underexplored.

Objectives

This cluster randomized controlled trial aimed to evaluate the effectiveness of a GP-delivered, group-based exercise intervention-designed based on the Theory of Planned Behavior (TPB) and operationalized with BCTs-on improving moderate-to-vigorous physical activity (MVPA) and physical performance among community-dwelling older adults.

Methods

A total of 429 older adults from 36 communities were randomized to either the intervention or control group. The randomization was performed at the community level. The primary outcome was MVPA, the secondary outcomes were timed up-and-go test time (TUG), Berg Balance Scale score (BBS), light intensity physical activity (LPA), falls efficacy, and loneliness. The intervention group participated in four weekly 60-minute group sessions that incorporated BCTs mapped to the TPB constructs, such as “modelling or demonstrating the behavior”, “social comparison”, “goal setting”, “behavioral contract”, etc. The control group received two monthly 60-minute group lectures designed to provide general health education on the knowledge of physical activity. Data were collected at baseline, 4 weeks (post-intervention) and 12 weeks. Adjusted difference was relied on linear mixed models.

Results

After 4 and 12 weeks, the intervention group demonstrated significant increases in MVPA compared to the control group, with a mean difference of 1.77 h/week (95% CI: 1.45, 2.10) and 1.96 h/week (95% CI: 1.64, 2.28), respectively. Significant improvements were also observed in BBS, TUG, and fall efficacy at 12 weeks. No significant between-group differences were found for LPA and loneliness.

Conclusions

The program could be suggested as an efficient intervention for improving physical performance and fall efficacy in community older adults due to the increase in PA.

Trial registration

International Traditional Medicine Clinical Trial Registry (ITMCTR2025002002). Retrospectively registered, with the registration dated 22/10/2025.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-026-07663-9.

Introduction

The World Health Organization’s (WHO) physical activity (PA) guidelines (2020) recommend that older adults aged 65 years and older should do muscle strengthening activities at moderate or greater intensity to improve balance and strength on 2 or more days a week [1]. However, a pooled analysis indicates that while the prevalence of insufficient physical activity remained stable or declined slightly until around age 60, it increased rapidly thereafter and was highest among the oldest age groups across all regions [2]. Cross-sectional data from low- and middle-income countries indicated that the prevalence of high leisure-time physical activity was 8.9% [3]. The data from the China Longitudinal Aging Social Survey (CHARLS) revealed that the average weekly duration of moderate-to-vigorous physical activity (MVPA) for people aged 59 and above was 48.17 min, which is lower than the recommended standards of WHO [4].

The factors influencing PA in older adults are multifaceted, among which interpersonal factors play a crucial role. External interpersonal interactions, such as social support, peer encouragement, and guidance from healthcare professionals, can effectively overcome inherent barriers to activity (e.g., lack of motivation, fear of risk) and enhance confidence and willingness to adhere to activity through the establishment of social connections [5, 6]. Therefore, building a supportive environment based on interpersonal interaction is a key point for designing effective interventions.

The Theory of Planned Behavior (TPB) posits that behavioral intention, shaped by attitude, subjective norms, and perceived behavioral control (PBC) [7–9]. The TPB has been widely applied in the field of PA promotion, both to predict PA behavior and to design interventions aimed at increasing PA [10–14]. Among these three constructs, subjective norm is an individual’s perception of social pressure from significant others (e.g., family, friends, healthcare providers) to engage in or refrain from a behavior, and directly addresses the interpersonal factors identified as crucial determinants of PA in older adult populations. This renders it highly suitable for designing group interaction-based community interventions, as it provides a theoretical framework for operationalizing social support mechanisms that can enhance intervention engagement and effectiveness. Furthermore, the PBC construct of TPB provides the direct theoretical mechanism to address the most salient age-specific barrier to PA-fear of falling, conceptually aligned with self-efficacy [8, 9], PBC captures individuals’ confidence in their ability to perform PA safely, which is a critical determinant of behavioral enactment in older adults. Besides, evidence supports that addressing attitudinal factors, specifically helping participants recognize that the benefits outweigh the barriers, is a viable pathway to behavior change [15].

Previous studies have demonstrated the feasibility of applying the TPB to PA interventions among older adults. For instance, a study by Stolte et al. [16] confirmed that TPB constructs strongly predict exercise intention among seniors, and their intervention successfully promoted PA. More recently, Qiao et al. [15] implemented a multi-theory intervention integrating TPB, which yielded significant and sustained improvements not only in frailty and physical function but also in cognitive and psychosocial outcomes (such as fear of falling, cognitive function, depressive symptoms, social support, etc.) for (pre) frail older adults.

However, based on the information provided, it would be challenging to operationalize these published studies. Therefore, systematic operational tools are needed to translate theory into practice. Behavior Change Techniques (BCTs) are a set of well-defined “active ingredients” used to alter determinants of behavior [8]. TPB offers a parsimonious and well-defined set of constructs, which conceptual clarity facilitates the systematic selection of corresponding BCTs, ensuring a transparent, replicable, and theoretically coherent intervention design tailored to address key determinants of PA initiation in older adults. A systematic review tests the efficacy of a TPB-based intervention but also contributes to methodological clarity by explicitly mapping and applying evidence-based BCTs to each TPB construct [17].

Against this background, this cluster randomized controlled trial (RCT) evaluated the effectiveness of a group-based physical activity intervention delivered by a general practitioner (GP) and grounded in the TPB, which was operationalized through specific BCTs. The results of this study will provide key evidence for the large-scale implementation of a theoretically grounded, clearly defined, and easily reproducible PA promotion model at the community level.

Method

Design

We conducted an open-label, cluster RCT with two parallel arms. The trial was conducted in 36 communities in Shanghai, China.

Study setting and participants

The study clusters were neighborhood or village committees within the Pudong New Area of Shanghai. To be eligible, a cluster must have had no existing physical activity programs for older adults and access to a suitable fixed venue.A research assistant, independent of the recruitment and intervention teams, generated the computer-based allocation sequence (Microsoft Excel) and randomly allocated the clusters in a 1:1 ratio to the intervention or control group.

Within clusters, individuals were eligible if they were permanent residents aged 60–85 years, did not meet WHO PA guidelines, and had impaired balance (X16 Balance Scale score ≤ 12) [18]. They also needed to be ambulatory and have sufficient cognitive/communication capacity for consent and group participation.

Exclusion criteria were: (1) acute illness or chronic disease exacerbation; (2) significant hearing, visual, or cognitive impairment; (3) history of neurological/psychiatric disorders; (4) major organ system disease contraindicating exercise; (5) vestibular/cerebellar or balance disorders; or (6) any other condition unsuitable for physical exercise.

Intervention

Participants in the intervention group received a GP-led, group-based exercise intervention, grounded in the TPB and operationalized through BCTs drawing on established classifications from the literature [19–22], which included four weekly on-site sessions (60 min each) and two follow-ups via telephone or WeChat. Each on-site session was structured as follows: a 20-minute health education component, a 30-minute supervised exercise practice, and a 10-minute period for structured social interaction. The content of these components was designed to target the key determinants of the TPB (attitude, subjective norms, and PBC), as well as the intention construct itself through specific behavior change techniques (e.g., goal setting, behavioral contract) that directly strengthen participants’ commitment to act. To further boost participant motivation and promote adherence, we incorporated an incentive strategy as a motivational component of the program.

The brief content of the intervention was as follows, the square brackets mean the BCT used in this intervention measure. Details of the intervention can be seen in Additional File 1.

Attitude

This included showing expert videos on the benefits of PA [Salience of consequences] and presenting a national policy briefing on free community PA facilities [Information about social & environmental consequences] to highlight the salience of consequences. Furthermore, participants engaged in exercises such as “Imagine being bedridden for your last decade—would you regret not acting sooner to stay active?” [Anticipated regret]. To foster cognitive restructuring, a pros-and-cons worksheet was used to list personal advantages and barriers followed by group problem-solving [Pros & cons].

Subjective norm

The intervention commenced with an official launch ceremony led by a Director of Preventive Healthcare at the Community Health Center to provide general social support [Social support general]. Practical social support was offered through the distribution of resistance bands, exercise logs, and class timetables [Social support practical]. During each session, the GP provided emotional social support by praising every participant by name [Social support emotional]. The modeling component involved the GP demonstrating three balance and strength tasks (e.g., one-leg stand for 10 s, weight-shift, and resistance-band chest press) after a standard instructional video, followed by paired peer correction [Demonstration of behavior]. To foster upward social comparison, a participant who performed the movements particularly well was selected each week to demonstrate the exercises [Social comparison].

Perceived behavioral control

To enhance perceived behavioral control, verbal persuasion about capability [Verbal persuasion about capability] was applied, wherein the GP recorded and provided individualized progress feedback to participants, such as noting, “You improved your one-leg stand from 3 seconds to 6 seconds, keep going!” Building on this, participants were assigned the homework “Recall one thing you stuck to,” with their experiences shared in the subsequent class to facilitate focus on past success [Focus on past success], thereby transferring mastery experiences. These strategies collectively supported the transition from intention to action by converting motivation into concrete plans.

Intention

Participants were asked to formulate a behavioral goal (e.g., “I will walk briskly for 30 min on Tuesday and Friday next week”) [Goal setting behavior] and a positive outcome goal (e.g., “In 12 weeks, I will be able to play with my grandchildren in the park for 30 minutes without feeling fatigued or short of breath.“) [Goal setting outcome]. Goal progression was reviewed and revised during every session and follow-up [Review behavior/outcome goals]. To enhance accountability, participants signed a behavioral contract regarding group attendance and homework adherence [Behavioral contract].

Incentive

Operational strategies were employed to boost motivation and adherence. Each participant was given two small practical gifts (e.g., tissues, towels, toothpaste, storage boxes) after every session [Material reward]; Furthermore, the promise of additional rewards for future attendance was used as an incentive [Material incentive]. These material incentives, although not a core component of the TPB model, were incorporated to minimize attrition.

Participants in the control group received a 45-minute health education on PA delivered by a GP every two weeks over a four-week period. The lecture covered foundational knowledge of PA, including different types of PA, the WHO’s recommended activity levels for older adults, and the benefits of regular exercise. No additional behavioral counseling, goal-setting, exercise sessions, or follow-up support was provided.

Process evaluation

The process evaluation was comprised of class attendance rates and participants’ satisfaction with the program. Participants’ satisfaction was assessed by the following 5 questions: Are you satisfied with the instructor’s explanation of exercise techniques? Do you gain any benefits from the group activities? Are you satisfied with the venue for the group activities? Are you satisfied with the organization and arrangement of the group activities?

Study outcomes

Physical activity (PA) level was assessed using two indicators: moderate-to-vigorous physical activity (MVPA), which served as the primary outcome, and light physical activity (LPA). The MVPA and LPA were measured using the Chinese version of the International Physical Activity Questionnaire Short Form (IPAQ-SF) [23]. The IPAQ-SF assesses PA over the past 7 days, including walking, moderate-intensity activities, and vigorous-intensity activities. Total physical activity was expressed as metabolic equivalents intensity (MET) minutes per week. Higher scores indicate greater levels of physical activity, representing a better outcome.

Physical function was assessed using three measures.Balance was assessed using the Timed Up and Go test (TUG) and the Berg Balance Scale (BBS) [24, 25]. The TUG measures the time (in seconds) required to stand up from a chair, walk 3 m, turn, walk back, and sit down; lower scores indicate better mobility. The BBS is a 14-item scale assessing static and dynamic balance, with scores ranging from 0 to 56; higher scores indicate better balance. In this sample, the BBS demonstrated excellent internal consistency (Cronbach’s α = 0.852), with factor analysis supporting its unidimensional structure (KMO = 0.898, Bartlett’s test χ² = 2058.939, p < 0.001).

Handgrip strength was measured in kilograms (kg) using a handgrip dynamometer (Jamar PLUS+). Participants performed two trials with each hand, and the maximum value was recorded. Higher scores indicate greater muscle strength.

Fear of falling was assessed using the Falls Efficacy Scale-International (FES-I), a 16-item instrument rated on a 4-point scale [26], a 16-item instrument rated on a 4-point scale (1 = not at all concerned to 4 = very concerned). Total scores range from 16 to 64, with higher scores indicating greater fear of falling (a worse outcome). In this sample, the FES-I exhibited excellent internal consistency (Cronbach’s α = 0.951), with adequate sampling adequacy (KMO = 0.936) and a significant Bartlett’s test of sphericity (χ² = 6717.997, p < 0.001).

Loneliness was assessed using the short-form UCLA Loneliness Scale (ULS-8), an 8-item instrument rated on a 4-point scale [27]. Total scores range from 8 to 32, with higher scores indicating greater loneliness (a worse outcome). In this sample, the ULS-8 showed modest internal consistency (Cronbach’s α = 0.686), which is consistent with values reported in prior studies involving Chinese community-dwelling older adults [28, 29]. The KMO value (0.774) and Bartlett’s test (χ² = 1054.667, p < 0.001) supported adequate construct validity.

Anthropometric and physiological measures were accessed by BMI, waist circumference, and blood pressure. BMI calculated as weight in kilograms divided by height in meters squared (kg/m²), with weight and height measured using a calibrated scale and stadiometer; waist circumference, measured at the midpoint between the lower rib and the iliac crest using a non-stretch tape; and resting blood pressure, measured in millimeters of mercury (mmHg) using a validated automatic sphygmomanometer.

The details of questionnaire of knowledge assessment are provided in Additional File 2.

Statistical analysis

The sample size was calculated a priori for the primary outcome MVPA using a standard formula for cluster randomized trials [28]. Assuming a two-sided α of 0.05, 90% power, and a standardized effect size of 0.635, we required 195 participants per group. This calculation was inflated by a design effect (DEFF) of 3.75 to account for clustering, based on an intra-cluster correlation (ICC) of 0.25 and an average cluster size of 12. To allow for an estimated 10% individual-level attrition, the final target sample size was set at 432 participants from 36 clusters. The sample size calculation formula is provided in Additional File 2.

All analyses followed the intention-to-treat principle, including all participants from randomized clusters. Descriptive statistics are presented as means and standard deviations (SDs) for normally distributed continuous variables, medians and interquartile ranges (IQRs) for skewed data, and counts and percentages for categorical variables. Baseline differences between groups were assessed using independent t-tests or chi-square tests, as appropriate.

The efficacy of the intervention was analyzed using linear mixed-effects models for continuous outcomes (MVPA, LPA, TUG, BBS, Fall efficacy, Loneliness), a random intercept for individual was included in all models. The model included fixed effects for intervention group, time, the group-by-time interaction, with adjustment for the prespecified covariates (age, sex, education, and chronic disease status). All other secondary outcomes were treated as exploratory. Additional File 3 presents the unadjusted descriptive statistics for all outcome measures, the between-group comparisons at each follow-up time point (4 and 12 weeks), and the results of the satisfaction evaluation.

Quality and fidelity

Rigorous quality control procedures were implemented to ensure data integrity and protocol adherence. A standardized training protocol was implemented for the GP who delivered the sessions in each cluster. The training, mandated for all GPs, consisted of three core components: (1) Theoretical instruction on the TPB and its application via specific BCTs; (2) Practical certification in the exercise protocol under the supervision of an exercise physiologist; and (3) Fidelity training on the consistent delivery of manualized content and use of session checklists. This process ensured that all facilitators possessed uniform, protocol-specific competency prior to intervention commencement. A dedicated quality supervisor, independent of the delivery team, periodically observed the intervention sessions using a structured checklist to monitor adherence to the protocol and ensure all core intervention elements were delivered as intended. Data collection forms were designed for completeness, and a double-data entry procedure was utilized. Outcome assessors were blinded to group allocation to minimize detection bias.

Results

Recruitment and participant characteristics

Participant recruitment took place from December 2022 to February 2023. A total of 576 individuals were screened, and 147 were excluded. The mean age of the included participants was 68.55 ± 5.50 years. The remaining 429 eligible participants were nested within communities that had been randomly allocated to the intervention and control groups. In the control group, two participants were lost to follow-up at the fourth week due to moving house (Fig. 1).

Fig. 1.

CONSORT flow diagram

Baseline demographic and sociological characteristics were compared between the intervention and control groups. There were no significant differences in gender, age, education level, and chronic disease group distributions. No significant between-group differences were found at baseline for BMI, waist circumference, DBP, grip strength, MVPA time, and fall efficacy. However, despite proper randomization, significant differences were noted for SBP, LPA time, TUG time, BBS score, and loneliness score, which can be attributed to random imbalance (Table 1).

Table 1.

Demographics and baseline characteristics

Variables	Intervention group (N = 215)/ n (%) or Mean (SD)	Control group (N = 214)/ n (%) or Mean (SD)	P value
Sex
Male	82 (38.14)	97 (45.33)	0.13
Female	133 (61.86)	117 (54.67)
Age
60–69 yr	128 (59.53)	137 (64.02)	0.34
70-yr	87 (40.47)	77 (35.98)
Education
< High school	170 (79.07)	168 (78.50)	0.89
≥High school	45 (20.93)	46 (21.50)
Chronic disease
Yes	133 (61.86)	139 (64.95)	0.51
No	82 (38.14)	75 (35.05)
BMI (kg/cm2)	24.81 (2.94)	25.29 (3.22)	0.14
Waist circumference (cm)	87.22 (8.93)	88.92 (9.74)	0.07
SBP (mmHg)	136.15 (17.00)	132.62 (13.73)	0.02
DBP (mmHg)	78.38 (8.82)	78.60 (8.62)	0.80
Grip (kg)	28.03 (8.92)	27.78 (8.57)	0.76
MVPA (hour/week)	0.04 (0.18)	0.04 (0.17)	0.96
LPA (hour/week)	4.39 (2.93)	5.12 (3.87)	< 0.01
TUG (s)	11.77 (3.77)	12.27 (3.60)	0.02
BBS (score)	49.44 (6.19)	50.95 (5.13)	0.01
Fall efficacy (score)	50.55(12.32)	52.61(10.42)	0.93
Loneliness (score)	13.53 (4.37)	12.24 (3.75)	<0.01

Abbreviations: SD Standard Error, BMI Body Mass Index, SBP Systolic Blood Pressure, DBP Diastolic Blood Pressure, MVPA Moderate-to-vigorous Physical Activity, LPA Light Physical Activity, TUG Timed Up and Go test, BBS Berg Balance Scale

Primary outcome

MVPA

After adjustment, the intervention group demonstrated a significant increase in MVPA from baseline, with an adjusted mean of 1.77 h/week (95% CI: 1.45, 2.10) at 4 weeks and 1.96 h/week (95% CI: 1.64, 2.28) at 12 weeks. The control group showed a smaller increase of 0.49 h/week (95% CI: 0.19, 0.79) at 4 weeks and 0.52 h/week (95% CI: 0.21, 0.82) at 12 weeks (Table 2; Fig. 2). The intervention led to a significantly greater increase in MVPA compared to the control group, with an adjusted mean difference of 1.28 h/week (95% CI: 0.94, 1.62) at 4 weeks and 1.44 h/week (95% CI: 1.10, 1.78) at 12 weeks (Table 3).

Table 2.

Within-group changes from baseline for all outcomes at 4 and 12 weeks

Variable	Intervention group/ Adjusted mean (95% CI)		Control group/ Adjusted mean (95% CI)
	4 weeks	12 weeks	4 weeks	12 weeks
Primary outcome
MVPA (hour/week)	1.77 (1.45, 2.10)***	1.96 (1.64, 2.28)***	0.49(0.19, 0.79)**	0.52 (0.21, 0.82)***
Secondary outcomes
LPA (hour/week)	0.80 (0.32, 1.27)**	0.78 (0.30, 1.26)**	0.46 (0.01, 0.92)*	0.36 (-0.10, 0.81)
TUG (s)	-1.14 (-1.59, -0.68)***	-1.56 (-2.01, -1.11)***	0.17 (-0.26, 0.60)	-0.44 (-0.87, -0.02)*
BBS (score)	2.27 (1.81, 2.72)***	3.74 (3.28, 4.19)***	0.38 (-0.05, 0.81)	0.62(0.19, 1.05)**
Fall efficacy (score)	2.96 (1.91, 4.00)***	5.56(4.52, 6.61)***	0.74 (-0.24, 1.73)	2.08 (1.09, 3.06)***
Loneliness (score)	-0.73 (-1.23, -0.23)**	-1.42(-1.92, -0.93)***	-0.43 (-0.90, 0.04)	-1.01(-1.48, -0.55)**

The adjusted mean within-group differences (95% CIs) were derived from linear mixed models, which included the intervention-by-time interaction, and were adjusted for age, sex, education, and chronic disease status, with a random subject intercept

Abbreviations: MVPA moderate-to-vigorous physical activity, LPA Light Physical Activity, TUG Timed up and go test, BBS Berg Balance Scale, CI Confidence Interval

* p < 0.05, ** p < 0.01, *** p < 0.001

Fig. 2.

Changes in Primary and Secondary Outcomes from Baseline at 4 and 12 Weeks (Adjusted Mean). Notes: Trajectories show the adjusted mean changes over time, derived from linear mixed models that accounted for the intervention effect over time and were adjusted for age, sex, education, and chronic disease status. Error bars indicate the standard error (SE).^* p < 0.05, ^** p < 0.01, ^*** p < 0.001. Abbreviations: MVPA, Moderate-to-vigorous Physical Activity; LPA, Light Physical Activity; TUG, Timed Up and Go test; BBS, Berg Balance Scale

Table 3.

Between-group change difference for all outcomes at 4 and 12 weeks

Variable	4weeks		12week
	Adjusted mean difference (95% CI)	P value	Adjusted mean difference (95% CI)	P value
Primary outcome
MVPA (h/w)	1.28 (0.94, 1.62)	< 0.001	1.44 (1.10, 1.78)	< 0.001
Secondary outcomes
LPA (h/w)	0.33 (-0.20, 0.86)	0.22	0.42 (-0.11, 0.96)	0.12
TUG (s)	-1.31 (-1.81, -0.80)	< 0.001	-1.12 (-1.62, -0.61)	< 0.001
BBS (score)	1.89 (1.39, 2.38)	< 0.001	3.12 (2.62, 3.61)	< 0.001
Fall efficacy (score)	2.21 (1.06, 3.37)	< 0.001	3.49 (2.33, 4.64)	< 0.001
Loneliness (score)	-0.30 (-0.84, 0.24)	0.27	-0.41 (-0.95, 0.13)	0.14

The adjusted mean between-group differences (95% CI) were derived from linear mixed models, which included the intervention-by-time interaction, and were adjusted for age, sex, education, and chronic disease status, with a random subject intercept

Abbreviations: MVPA Moderate-to-vigorous Physical Activity, LPA Light Physical Activity, TUG Timed Up and Go test, BBS Berg Balance Scale, CI Confidence Interval

Secondary outcomes

LPA

Both groups showed a significant increase in LPA at 4 weeks and 12 weeks (Table 2; Fig. 2). However, for LPA, no statistically significant between-group differences were found at either time point, with a mean difference of 0.33 h per week (95% CI: -0.20, 0.86; p = 0.12) at 4 weeks and 0.42 h per week (95% CI: -0.11, 0.96; p = 0.22) at 12 weeks (Table 3).

Balance function

The TUG test performance improved significantly more in the intervention group than in the control group, with a mean difference of -1.31 s (95% CI: -1.81, -0.80; p < 0.001) at 4 weeks and - 1.12 s (95% CI: -1.62, -0.61; p < 0.001) at 12 weeks. Similarly, BBS scores demonstrated significantly greater improvement in the intervention group, showing a mean difference of 1.89 points (95% CI: 1.39, 2.38; p < 0.001) at 4 weeks and 3.12 points (95% CI: 2.62, 3.61; p < 0.001) at 12 weeks (Table 3). Additional correlation analysis between MVPA and physical performance (TUG, BBS) is provided in Additional File 3.

Fall efficacy

Fall Efficacy scores improved significantly more for the intervention group, with between-group differences of 2.21 points (95% CI: 1.06, 3.37; p < 0.001) at 4 weeks and 3.49 points (95% CI: 2.33, 4.64; p < 0.001) at 12 weeks (Table 3).

Loneliness

Both groups showed a reduction in loneliness scores at 12 weeks. The difference in improvement between the groups was not statistically significant at either 4 weeks (mean difference: -0.30, 95% CI: -0.84, 0.24; p = 0.14) or 12 weeks (mean difference: -0.41, 95% CI: -0.95, 0.13; p = 0.27). Table 2 and Fig. 2 shows within-group changes from baseline for all outcomes at 4 and 12 weeks.

Process evaluation

The attendance rate and follow-up rate of the research subjects in the intervention group for health education activities were both higher than 85%, and the overall satisfaction score was higher than 4.8 points (with a maximum of 5 points). Satisfaction score of each item can be seen in Additonal File 3.

Discussion

This 4-week, GP-delivered, theory-based exercise intervention successfully increased PA and improved physical function in community-dwelling older adults. The intervention group demonstrated a significantly greater increase in MVPA compared to the control group at the 12-week follow-up. Concurrently, significant improvements were observed in TUG, BBS, and fall-related self-efficacy, indicating that the intervention had a positive effect across multiple domains of physical function.

Previous studies employing the TPB framework have successfully increased PA in various adult populations. For instance, Gholamnia-Shirvani et al. [29] conducted a multimedia-based health education RCT grounded in an extended TPB model among 130 military wives in Tehran, Iran, and observed significant improvements in PA and fitness in the intervention group after 6 months. Similarly, Huang et al. [10] implemented a worksite intervention where a TPB-based tailored Short Message Service (SMS) program combined with web-based knowledge led to significantly higher PA levels compared to web-based knowledge alone after 12 weeks. Building upon this established evidence base for theory-driven physical activity interventions, the result of the present study extends the literature by operationalizing the TPB through a general practitioner-led, group-based format specifically tailored for community-dwelling older adults. In addition to the increase in PA in the intervention group, interestingly, a slight increase in PA was also noted in the control group compared to baseline. This may be attributed to the general health education they received, which highlighted the benefits of activities like Tai Chi and Baduanjin, or to the Hawthorne effect arising from repeated PA measurements, a phenomenon supported by Freene et al. [30], who reported PA increases simply through frequent measurement and brief advice.Although our findings align with previous TPB‑based interventions in demonstrating positive effects on PA, the magnitude and pattern of effects observed in our study may differ from prior work for reasons discussed in the following paragraphs, including features of our intervention model and the unique characteristics of our sample.

An intriguing finding was the greater increase in MVPA than in LPA in our intervention group, which contrasts with some previous reports, such as Sanaeinasab et al. [31], which reported smaller changes in vigorous PA relative to light or moderate PA. MVPA and LPA are based on MET. We propose that this discrepancy reflects the design of our intervention. Unlike programs focused primarily on walking or general aerobic activity, our protocol included structured, supervised, and progressive strength and balance training exercises. These activities require effort levels consistent with moderate-to-vigorous intensity and were likely perceived and reported by participants as MVPA. This may explain the predominant increase in MVPA observed in our study.

At 12 weeks, the adjusted between-group differences significantly favored the intervention group, with a reduction of 1.12 s in the TUG test time, an increase of 3.12 points on the BBS. However, the magnitude of these improvements was smaller than the effects reported in some previous studies [32–38]. For example, Leem et al. [37] reported a 4.08 s reduction in TUG test time (from a baseline of 16.17 s) following a 12-week Otago program combined with action observation training (3 sessions/week, 20 min/session), and a 3.76 s reduction (from a baseline of 16.56 s) with the Otago exercise alone (3 sessions/week, 50 min/session). Jahanpeyma et al. [38] reported a 6-point increase in the BBS score from baseline of 44 points after a 12-week Otago exercise program (3 sessions/week, 45 min/session). A network meta-analysis showed that combined exercise interventions (e.g., resistance plus aerobic, or resistance plus balance training) resulted in a reduction of 0.83 to 1.85 s in the TUG test time among older adults. The TUG improvement seen in our trial (1.31 s) is congruent with this reported efficacy range [39].Consistent with previous research, our findings improved the balance function of the participants; however, the magnitude of improvement was modest compared to some prior studies. A possible explanation is that participants in this study had relatively good balance, strength, and high fall efficacy at baseline. This may be attributable, at least in part, to the use of the X16 scale as a screening tool for inclusion, rather than more sensitive measures (BBS and TUG test), which may have selected a higher-functioning sample. Additionally, the intervention duration was shorter than that in the studies mentioned above.

Despite a higher baseline, fall efficacy (on the FES-I scale) still improved significantly in the intervention group compared to the control group, with a between-group difference of 3.49 points at 12 weeks. This finding is consistent with other reports. For example, Parry et al. [40] found a significant improvement in fall efficacy (a reduction of -4.02 points; 95% CI: -5.95 to -2.1) following a cognitive-based physical activity intervention among 415 older adults. FilipoviĆ et al. [41] found that improvements were detected with TUG, Sit To Stand test (STS), FES-I after 12 weeks of exercise intervention. The improvement in fall efficacy likely resulted from the intervention acting through both physical gains in function and balance, and by the group setting building psychological confidence via successful experiences. This interpretation is supported by qualitative evidence showing that, among individuals with physical vulnerability, regular exercisers perceive less hazard in physical activity than non-exercisers, and that positive exercise experiences in a supportive social context reshape risk appraisals and build confidence [42].

Both groups showed significant improvements in loneliness from baseline to follow-up, yet no significant between-group difference was observed. This result may be explained by two interrelated factors. First, the intervention group participated in a structured, TPB-based PA program delivered in a group setting, which contains the BCTs target social norm including Social support, Demonstration of the behaviour, and Social comparison, which may directly result in both the improvement of loneliness and MVPA [5, 6, 43]. While the control group also received group-based health education lectures to had opportunities for social interaction (in-person contact), factors known to reduce loneliness [43]. Second, increased PA can indirectly reduce loneliness by facilitating social interaction and enhancing psychological well-being, these effects may take longer to emerge than improvements in physical activity itself, the 12-week duration may have been insufficient to alter a complex psychosocial state like loneliness, or the measurement tool may have lacked sensitivity to detect subtle changes [44]. Although the intervention group achieved significantly greater improvements in MVPA compared to the control group, the 4-week intervention period may not have been sufficient for these behavioral changes to translate into meaningful reductions in loneliness. It is worth noting that while a significant within-group improvement was observed in the intervention group, this finding is consistent with the results of previous related PA intervention studies [45]. The change was not significantly greater than that observed in the control group. Such a group-based format could easily be categorized as a form of low-intensity social intervention [46], the reduction in loneliness within the control group likely represents a non-specific therapeutic effect of regular group socialization.

The differential findings between MVPA and loneliness further underscore the complex role of interpersonal factors in intervention outcomes. As discussed in the Introduction, interpersonal factors are critical determinants of PA in older adults. The significant improvement in MVPA in the intervention group suggests that the group-based format effectively harnessed social support and peer modeling to promote behavior change. However, loneliness showed only within-group improvements with no significant between-group difference. This pattern may reflect that while the intervention provided structured social interaction sufficient to influence PA, a behavior that can change relatively quickly, it may not have sustained the depth or duration of social connection required to meaningfully reduce loneliness. This interpretation suggests that interpersonal factors may operate through different mechanisms or time courses for different outcomes. Future interventions may need to explicitly target social bonding in addition to social support to achieve broader psychosocial benefits.

Limitations

First, the reliance on self-reported physical activity data via the IPAQ-SF is a notable limitation, as it is susceptible to recall and social desirability biases. This may have led to an over- or under-estimation of true physical activity levels. Future studies would benefit from incorporating objective measures, such as accelerometers or heart rate monitors, to validate and complement self-reported data. Second, the relatively brief duration of the intensive intervention and the short-term follow-up limit the assessment of the long-term efficacy and sustainability of the intervention. It remains unclear whether the observed significant improvements in MVPA, balance, and fall efficacy would be maintained over a longer period (e.g., 6 or 12 months). Future research should implement longer intervention and follow-up periods to evaluate the persistence of effects. Third, the generalizability of the findings may be constrained by the relatively high baseline functional status of the recruited participants. Their near-normal levels of balance, strength, and fall efficacy at baseline potentially created a ceiling effect, limiting the magnitude of observable improvement and possibly attenuating the estimated effect size of the intervention. Consequently, the effectiveness of this intervention in frail older adults or those with significant functional impairments requires further investigation. Forth, while a pilot study informed protocol refinement, we did not conduct a formal pre-delivery fidelity assessment. Thus, undetected protocol deviations may have occurred, introducing delivery variability and potentially biasing our effect estimate toward the null. Fifth, the intervention was delivered by general practitioners within a specific community setting in Shanghai. This may limit the generalizability of the findings to other healthcare systems, cultural contexts, or delivery agents (e.g., nurses, physiotherapists, or community volunteers). Future studies should test the effectiveness and feasibility of this model across diverse settings and populations.

Implications

The findings offer preliminary insights for future research. For research, future studies should: (1) conduct mediation analyses to formally test the proposed TPB mechanisms; (2) pursue longer-term effectiveness trials (≥ 6 months) to assess sustainability; (3) explore intervention optimization, for example, through personalization (tailoring BCTs to individual barriers or preferences) or digital augmentation (using mobile apps or messaging to extend support and reduce resource intensity), these strategies aim to address heterogeneity in response and improve long-term adherence; and (4) perform economic evaluations to assess cost-effectiveness. For practice, the results suggest that: (1) integration into primary care via GPs is feasible, though scalability requires further testing; (2) skills- and confidence-building techniques may be beneficial; and (3) the group format may foster social support and normative influence. All these implications should be validated in larger, longer, and more diverse samples before widespread implementation.

Conclusion

A theory-driven (TPB-based) exercise intervention, operationalized with specific BCTs and delivered by GP, significantly increased MVPA and improved balance function and fall efficacy in community-dwelling older adults. This structured, model demonstrates the potential for integrating theory into primary care practice to promote healthy aging at scale.

Abbreviations

BBS

Berg Balance Scale

BCTs

Behavior Change Techniques

BMI

Body Mass Index

CI

Confidence Interval

DBP

Diastolic Blood Pressure

DEFF

Design Effect

FES-I

Falls Efficacy Scale International

GP

General Practitioner

RCT

Randomized Controlled Trial

ICC

Intra-cluster Correlation

IPAQ-SF

International Physical Activity Questionnaire Short Form

IQRs

Interquartile Ranges

LPA

Light Physical Activity

MVPA

Moderate-to-vigorous Physical Activity

MET

Metabolic Equivalent Intensity

SBP

Systolic Blood Pressure

SE

Standard Error

SMS

Short Message Service

STS

Sit to Stand test

TPB

Theory of Planned Behavior

TUG

Timed Up and Go test

Authors' contributions

JLY supervised the data collection, drafted the original version of the manuscript, and edited the final version; YYS conducted the statistical analysis, assisted with data collection, and collaborated on the manuscript draft; LSJ assisted with data collection and collaborated on the manuscript draft; CY collaborated on the manuscript draft; YNJ contributed to the study design and collaborated on the manuscript draft; PPZ conceived the study, contributed to the study design, and collaborated on the manuscript draft; LZ conceived the study, provided overall guidance and supervision, and collaborated on the manuscript draft. All authors read and approved the final manuscript.

Funding

This work was funded by the National Natural Science Foundation of China (NSFC) [Grant Number 72474052], the Young Medical Talents Training Program of Shanghai Pudong Health Commission [grant number PWRq2024-19] and Shanghai Pudong New Area Center for Disease Control and Prevention [grant number PDCDC-KJ-2024-20].

Data availability

The data presented in this study are available on request from the corresponding author.

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of the Shanghai Pudong New Area Center for Disease Control and Prevention (Approval Number: PDCDCLL-20230420-001). All procedures and methods performed in the study were in accordance with the Declaration of Helsinki. Informed consent has been collected from all enrolled participants. This trial was retrospectively registered with the International Traditional Medicine Clinical Trial Registry (ITMCTR) on 22 October 2025 (Registration Number: ITMCTR2025002002). The CONSORT checklist can be seen in Additional File 4.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.