Predictors of real‐world adherence to prescribed home exercise in older patients with a risk of falling: A prospective observational study

Abstract Objectives Using a multi‐ethnic Asian population, this study assessed adherence to prescribed home exercise programs, explored factors predicting adherence, and evaluated whether home exercise adherence was associated with physical activity. Methods A prospective cohort study was conducted in 68 older adults (aged ≥65 years) from two geriatric outpatient clinics in Singapore, who were receiving tailored home exercises while undergoing 6 weeks of outpatient physical therapy for falls prevention. Adherence was measured as the percentage of prescribed sessions completed. Predictor variables included sociodemographic factors, clinical characteristics, intervention‐specific factors, and physical and psychosocial measures. Multivariable linear regressions were performed to develop a model that best predicted adherence to prescribed exercise. Physical activity levels, measured by accelerometry, were analyzed by cross‐sectional univariate analysis at 6 weeks. Results The mean adherence rate was 65% (SD 34.3%). In the regression model, the number of medications [B = 0.360, 95% CI (0.098–0.630)], social support for exercising [B = 0.080, 95% CI (0.015–0.145)], and self‐efficacy for exercising [B = −0.034, 95% CI (−0.068–0.000)] significantly explained 31% (R 2 = 0.312) of the variance in exercise adherence. Older adults with better adherence took more steps/day at 6 weeks [B = 0.001, 95% CI (0.000–0.001)]. Conclusions Low adherence to home exercise programs among older adults in Singapore, emphasizing the need for improvement. Counterintuitively, older adults with more medications, lower exercise self‐efficacy, but with greater social support demonstrated higher adherence. Addressing unmet social support needs is crucial for enhancing adherence rates and reducing fall risks.


| INTRODUC TI ON
It is well established that home exercise programs can reduce falls rate and falls risk. 1 Studies have shown that higher doses of exercise, specifically 3+ h per week of balance-specific training, can have greater effect on reducing the rate of falls in older adults at risk of falling. 2 One common approach to ensure sufficient dose is to prescribe unsupervised home exercise; however, achieving adherence to home exercise can be challenging.Adherence refers to the extent to which an individual's behavior corresponds with agreed recommendations from a health care provider. 3Non-adherence to treatment leads to more frequent falls, greater use of health care resources from falls-related injuries, 4 and less success in decreasing falls. 5e efficacy of individualized home-based exercise has been clearly demonstrated in randomized controlled trials. 1,6However, adherence to prescribed exercise in real-world settings has not been well characterized.Reported adherence to prescribed home-based exercise in the literature has varied, with one review reporting a rate of 21%, 7 and is known to decline over time. 8While there is no consensus on the definition of adequate adherence at present, 9 overall adherence is typically inadequate.
Limited and inconsistent findings exist regarding predictors of adherence to falls-prevention home exercises in older adults.One study found that taking fewer medications was associated with better adherence. 10Another study showed that better cognitive function and functional mobility predicted greater adherence to the Otago Exercise Program. 11Additionally, higher levels of physical activity predicted increased exercise adherence. 12Understanding the association between adherence to home exercise programs and subsequent physical activity levels is important. 6ven the limited information on adherence to prescribed exercise, the present study aimed to: (1) evaluate real-world adherence to prescribed home exercise; (2) explore factors predicting adherence to prescribed home exercise programs designed to reduce falls risk/falls rate in older adults during the initial 6 weeks of rehabilitation; and (3) determine any association between adherence to home exercise and physical activity levels at 6 weeks.

| Adherence measure
Participants were asked to complete a paper-based exercise diary recording each time they completed a session of their individually prescribed home exercise program over the 6-week treatment period.Adherence was defined as percent completion of prescribed home exercises over 6 weeks, following the commonly reported definition in the literature. 14Adherence rate was calculated by way of the total number of home exercise sessions completed as a proportion of that prescribed each week from T0 to T6, to determine each participant's overall completion rate (%), and completion rate per week (%).

| Potential predictors
Selection of potential predictors was informed by previous literature and constructs of the COM-B behavior change model. 15Thirty-one variables were selected (Appendix S1).Sociodemographic and clinical characteristics were collected through a self-report questionnaire administered.Intervention-specific variables were obtained from physical therapy records, including home exercise prescription details and follow-up information.Physical and psychological abilities were measured using validated measurements such as the 30-s chair stand test, 16 the Timed Up and Go test, 17 the Falls Efficacy Scale-International, 18,19 and the Self-Efficacy for Exercise Scale. 20cial opportunity influences were measured using the validated 13-item Social Support for Exercise Behaviors survey that assesses perceived social support for exercise during the past 3 months from family and friends, 21 and a 3-item short form of the Revised University of California Los Angeles Loneliness Scale was used.22 Physical activity was measured using the Actigraph GT3X accelerometer (Ac-tiGraph) that was worn at T6 (6 weeks from baseline) and the Phone-FITT survey.

| Data analysis
Baseline characteristics were analyzed using descriptive statistics.
Categorical variables were described as percentages and continuous variables as mean (SD).It has been proposed that selecting variables begins with univariate analyses 24 ; we assessed all 31 potential predictors (Appendix S1) using univariate linear regression.
Potential predictor variables with a univariate association with the primary outcome at a significance level of p < 0.15 were retained as candidate predictor variables for the multivariate linear regression model.A higher significance level (p < 0.15 compared to p < 0.05) was used to select variables so that important variables relevant to the outcome were not missed, and to avoid deleting less significant variables that had practical or clinical significance. 25To obtain the final multivariable linear regression model for adherence, the alpha level was set at 0.05 and backward elimination was used.Percent exercise adherence data were not normally distributed and were therefore transformed using a square root transformation.Preliminary analyses were performed to ensure there were no violations of the assumptions of normality, linearity, homoscedasticity, and multicollinearity.Cases were excluded if standardized residuals exceeded ±2.SPSS (v26; IBM Corp) was used for all statistical analyses.We conducted a cross-sectional analysis for physical activity measured by accelerometer at T6 using univariate linear regression.
Minimum wear time was defined as 10 h per day, and a minimum number of 4 valid days (with at least 1 weekend day) was required for inclusion in data analysis. 26Non-wear time was considered as 60 min of consecutive zeros.Total step counts (average steps/day) and minutes (min/day) spent in sedentary, light, moderate, and vigorous physical activity were derived using cut points (counts/minute) for: sedentary 0-99; light ≥100-2019; moderate 2020-5998; vigorous ≥5998; moderate-to-vigorous intensity ≥2020 activity. 27,28

| Participants
Seventy-two patients were initially recruited, but four participants withdrew prior to T0.However, data collection was ceased in March 2020 due to COVID-19 restrictions.As a result, data analysis was conducted for 68 participants (n = 68; 78 ± 6.7 years) (Table 1).

| Adherence rate
Out of the initial 68 participants, 60 were included in the analysis, as 8 participants did not return their exercise diary.The mean adherence to the 6-week home exercise program was 65% TA B L E 1 Characteristics of the study sample.

Sociodemographic characteristics n Value
Age (years) 78.9 ± 6. (SD: 34.3%).Among the participants, 12 (20%) were fully adherent, completing all prescribed exercise sessions while 5 (8%) did not initiate or perform the prescribed exercises.Notably, the mean adherence rate exhibited a downward trend over the 6-week period (Figure 1), and Figure 2 illustrates the increasing number of participants with non-adherence as the weeks progressed and vice versa.The final multivariable regression model (Table 3) included three significant predictors (number of medications, social support for exercise, and self-efficacy for exercise) that explained 31% (R 2 = 0.312) of the variance in exercise adherence F (3, 40) = 6.045, p = 0.002 (Table 3).

| Predictive factors
A comparison was made between baseline data for those participants included (n = 44) and not included (n = 16) in the final model.No statistically significant differences were found between groups with respect to characteristics in Table 1 (p ≤ 0.05) (Appendix S2).

| Physical activity
Out of 68 participants, valid accelerometry data were available for 48 participants.Fourteen participants had missing data for various reasons: eight did not complete the study, three misplaced their accelerometer, three refused to wear the accelerometer, and six did not meet minimum wear time criteria.The physical activity and sedentary behavior of the 48 participants at T6 are presented in Table 4.
In the cross-sectional univariate analysis at T6, average steps per day within the physical activity factors was found to be significantly associated with exercise adherence.This association explained a significant proportion of variance in adherence scores, with an R 2 value of 0.132, F (1, 45) = 6.848, p = 0.012.Adherence rate (%)

Week
Adherence to prescribed home-based exercise over 6 weeks we observed a decline in adherence over time, consistent with trends seen in older adults with initially high adherence rates. 8This suggests the need for effective strategies to monitor and support exercise adherence in older adults, coupled with early intervention and collaborative efforts to overcome barriers and promote successful outcomes.
At baseline, we explored several possible predictive measures, but these variables were not able to explain most of the variability (69%) in exercise adherence.It is likely that unknown factors play a greater role in the prediction of exercise adherence in older adults at risk of falling and were not measured in this study.Factors such as depression, 29 cognitive function (e.g., executive function), 11 and health status 30 were not measured in the current study but have previously been related to exercise adherence.The use of self-report for measuring exercise adherence introduces potential variability in the accuracy of the dependent variable data, highlighting the complex and multifactorial nature of exercise adherence.
Contrary to previous studies, 10,31,32 baseline medication use predicted positive home exercise adherence, possibly due to differences in study design, sample characteristics, and attitude towards health instructions. 10Tailoring exercise programs to specific populations is crucial due to adherence complexity.Further investigation is needed to explore the influence of medication count on exercise adherence and its impact on physical activity levels, considering medications such as benzodiazepines and anticholinergics that hinder functional status in older adults. 33Yet, enhancing exercise capability may not translate to increased physical activity levels. 34Motivation for health improvement may vary based on medication usage, with individuals taking fewer medications showing less concern, while those with multiple medications making greater efforts, aligning with the Health Belief Model. 35herence to prescribed exercise was positively associated with social support, aligning with the concept that social support can enhance exercise behavior by strengthening self-efficacy and expectations regarding exercise benefits. 36This finding is consistent with previous review on adherence to home-based exercise and social support. 37However, it is concerning that the average scores for social support in our study (family: 32%, friends: 26%) were relatively low.Despite a high proportion of older adults living with their families in multigeneration households, they still lacked sufficient individual-level support to adhere to home exercise programs.The absence of, or poor social support for exercise requires further investigation.It is possible that limited awareness of exercise benefits for fall prevention and reduced concern about falls during exercise may have inadvertently constrained their activities. 36Further, family members in Singapore may have limited time to supervise exercise, due to other work or social obligations (i.e., needing to provide care to both the young and older generations). 38Hence, families need to be better equipped to support and complement caregivers/ health professionals to facilitate exercise adherence.
[41] This could be due to overestimation of task demands when requirements are unclear, while underestimation leads to underconfidence. 42Ambiguity about the nature of the activity moderates the self-efficacy-performance relationship, with a negative relation at high ambiguity and a positive relation at low ambiguity. 43High selfefficacy individuals may allocate fewer resources, decreasing performance. 44Low self-efficacy individuals may receive more social support for exercise, positively impacting adherence.[47] Healthcare providers should provide information, education, 48  Gender, level of education, and performance on the 30 s chair stand test did not appear to predict exercise adherence.This result is surprising, given that previous studies reported that older adults who were female, 49 with higher education levels (>high school), 50 and poor functional performance, 11 were more likely to engage with/adhere to exercise activities to prevent falls.Further investigation is required to determine the relative importance of these and other potential predictors of adherence that provide valuable information about which patients are more/less likely to adhere to falls prevention exercise programs, and thereafter maintain ongoing participation in exercise.
Higher exercise adherence correlates with increased daily step count, suggesting a potential connection, but causality cannot be inferred due to the study's cross-sectional design.Prior research indicates that low adherence hampers exercise effectiveness, impacting physical function/performance. 51 As our participants were clinical populations with low physical activity levels, they would benefit most from increasing exercise.Clinicians should offer optimal guidance, including setting exercise goals, to encourage ongoing participation.

| Limitations
This study has several limitations.Firstly, the main outcome relied on self-report exercise diaries, which may introduce bias towards over-reporting. 52,53However, diaries offer valuable information on exercise type, individual sessions, and contextual factors that accelerometry cannot capture.Additionally, self-report diaries are simple and cost-effective methods for prospective data collection. 54

ACK N O WLE D G E M ENTS
The authors acknowledge the support of their clinical colleagues and their department heads at the two facilities at which data were collected.In particular, the authors are grateful to Ms Toylyn Lee, Ms Faezah Ghazali, and Ms Melissa Heng who were the liaison persons for their respective facilities and all the research participants who participated in this study.

FU N D I N G I N FO R M ATI O N
Not applicable.

CO N FLI C T O F I NTER E S T S TATEM ENT
None declared.

E TH I C S S TATEM ENT
Ethic approval for the study was gained from The Singapore National Healthcare Group -Domain Specific Review Board (2018/01372) and The University of Queensland Human Research Ethics Committee (HREC/2019001991).

CO N S E NT
Participants provided consent for the anonymous publication of data.

2. 1
| Design An observational cohort study was conducted between May 2019 and March 2020 with assessment time points of T0 (baseline) and T6 (6 week follow up).Participants were recruited through convenience sampling from two major hospital-based geriatric outpatient clinics in Singapore.Participants were involved in a 6-week, individually tailored home exercise intervention while undergoing outpatient physical therapy for falls prevention delivered in public hospitals.Ethical approval was obtained from The Singapore National Healthcare Group -Domain Specific Review Board (2018/01372) and The University of Queensland Human Research Ethics Committee (HREC/2019001991).All participants provided written informed consent prior to enrolment in the study.2.2 | Participants and settingParticipants aged 65 years or above, deemed at increased risk of falls, were included in the study.Increased risk was defined by attendance at an outpatient falls and balance clinic (with/without history of falls) and assessment by a geriatrician and/or physiotherapist in hospital-based geriatric outpatient clinics.Exclusion criteria comprised nursing home residency and cognitive impairment hindering the ability to provide informed consent or understand study procedures.Cognitive status was evaluated during initial screening using the Short Portable Mental Status Questionnaire, considering cognitive impairment if there were ≥3 errors.13

23
Secondly, caution is warranted in interpreting the findings due to the small sample size.Recruitment during the COVID-19 pandemic and TA B L E 2 Univariate linear regression between exercise adherence and sociodemographic, clinical, intervention-specific, physical, psychosocial variables (n = 60 for all variables except 30-s chair stand, n = 45, and timed up and go, n = 47).

TA B L E 2
(Continued)  restrictions on community activities (such as research) in Singapore limited participation.This small sample size may have prevented us from detecting some significant relationships.Therefore, we acknowledge the potential for chance findings and emphasize the need for a cautious interpretation of the results.Thirdly, convenience sampling introduces the possibility of selection bias.The findings are limited to older adults at risk of falling and may not be extrapolated to other populations (e.g., institutionalized, healthy people).This study provides evidence for the real-world adherence patterns of older adults at risk of falling undergoing standard care at hospital-based outpatient clinics as well as factors predicting home exercise adherence.Older adults in Singapore who took more medications and had better social support, but low self-efficacy, were most likely to adhere to prescribed home exercise programs for falls prevention.Various factors, including individual characteristics (e.g., medication intake, use of a walking aid), program features (e.g., functional exercises), physical capability (e.g., strength), and psychosocial qualities (e.g., self-efficacy, social support) may serve as indicators of poor exercise adherence.Further work is needed to explore the interaction of these factors and other predictors to enhance adherence to exercise recommendations.This calls for a larger sample size to corroborate and broaden our findings, alongside a recognition of the significance of replication studies in validating our outcomes.AUTH O R CO NTR I B UTI O N SConcept and design: Bernadine Teng, Sandra G. Brauer, Anna L. Hatton, and Sjaan R. Gomersall.Data collection: Bernadine Teng.Data analysis: Bernadine Teng, Sandra G. Brauer, Asaduzzaman Khan, and Sjaan R. Gomersall.All authors reviewed and gave approval for the final version.